Non-line-of-sight moving target three-dimensional reconstruction method based on MIMO millimeter wave radar

Abstract

The invention discloses a non-line-of-sight moving target three-dimensional reconstruction method based on an MIMO millimeter wave radar, and is applied to the technical field of radar target detection in a complex urban environment. Aiming at the problem of difficulty in realizing three-dimensional reconstruction of a non-line-of-sight moving target in the existing radar target detection technology, the method comprises the following steps: firstly, establishing a non-line-of-sight multipath propagation model, and exporting a multipath signal model of an MIMO millimeter wave radar; the invention further provides a multipath signal departure angle (DoD) / arrival angle (DoA) unambiguous estimation algorithm, and solves the problem that a main lobe and a grating lobe are difficult to distinguish in the DoD / DoA estimation process of an existing commercial MIMO radar. And finally, proposing a path-oriented non-line-of-sight target three-dimensional positioning algorithm, respectively positioning the non-line-of-sight target by using different paths, and fusing different path positioning results to form a three-dimensional point cloud. According to the method, three-dimensional reconstruction of a plurality of non-line-of-sight moving targets can be realized, and the correlation problem between paths does not need to be considered.

Description

Technical Field

[0001] This invention belongs to the field of non-line-of-sight target detection technology, and specifically relates to a non-line-of-sight moving target three-dimensional reconstruction method based on MIMO millimeter-wave radar. Background Technology

[0002] Non-line-of-sight (NLOS) target detection technology locates obscured targets by inverting multipath signals generated by the reflection and diffraction of electromagnetic waves on building surfaces, which is beneficial to improving the early warning performance of radar systems. Unlike the unique mapping relationship between traditional radar echoes and targets, multipath echoes are strongly coupled and difficult to separate. Therefore, multipath identification has become the key and prerequisite for NLOS target localization, and many research institutions at home and abroad have conducted extensive research on this issue. The literature "Looking behind acorner using multipath-exploiting UWB radar, IEEE Trans. Aerosp. Electron.Syst., vol. 51, no. 3, pp. 1916–1926, Jul. 2015" uses range measurement of a single-transmitter, single-receiver radar to identify multipath echoes of NLOS targets, and then uses the elliptical cross-location method to derive the position of the NLOS target. However, this method can only locate a single target. The paper "Multi-Domain Features-Based NLOSTarget Localization Method for MIMO UWB Radar, IEEE Sensors J., vol. 23, no.23, pp. 29314-29322, Dec. 2023" proposes using the range-Doppler topological features of multipath signals to achieve multipath identification and thus locate multiple non-line-of-sight targets. However, the above methods all require the completeness of the multipath signals so that the multipath signals exhibit clear features in dimensions such as range and Doppler. When some paths are missing, it is difficult to ensure the usability of the algorithm.

[0003] MIMO radar, by transmitting orthogonal waveforms, grants itself additional degrees of freedom at the transmitting end. It has been proven to accurately identify multipath signals by distinguishing between the direction of departure (DoD) and direction of arrival (DoA) of the received signal. However, the element spacing of existing co-located MIMO radar transmit arrays is often several times half a wavelength, making it difficult to achieve unambiguous estimation of the DoD / DoA of multipath signals. Furthermore, most existing methods can only achieve two-dimensional reconstruction of non-line-of-sight targets, failing to reflect the target's attitude information. Therefore, solving the problem of unambiguous estimation of DoD / DoA in co-located MIMO radar, and further realizing three-dimensional reconstruction of non-line-of-sight moving targets, has significant practical implications. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a non-line-of-sight moving target 3D reconstruction method based on the high-resolution characteristics of millimeter-wave radar. This method first performs unambiguous estimation of the DoD / DoA of multipath signals, determines the type of multipath signal based on the difference between the two, and then uses a path-guided strategy to achieve 3D localization of non-line-of-sight targets. By fusing point clouds from different paths, the 3D reconstruction of non-line-of-sight targets is achieved.

[0005] The technical solution adopted in this invention is: a non-line-of-sight moving target 3D reconstruction method based on MIMO millimeter-wave radar, comprising:

[0006] S1. Construct a non-line-of-sight multipath propagation model and derive the multipath signal model of MIMO millimeter-wave radar;

[0007] S2. Perform range-Doppler processing on the received signal represented by the multipath signal model in step S1, and detect non-line-of-sight moving targets;

[0008] S3. Based on the detection results of S2, the DoD / DoA of the multipath signal is first coarsely estimated by utilizing the symmetric feature of the main lobe about the spectral diagonal in the two-dimensional DoD / DoA spectrum of the multipath signal. Then, the hypothesis testing method is used to achieve an unambiguous estimation of the DoD / DoA of the multipath signal.

[0009] S4. Based on the DoD / DoA estimated in S3, combined with the non-line-of-sight multipath propagation model established in S1, the path... , and The target is located, and the point cloud data from the three sources is integrated to achieve 3D reconstruction of the non-line-of-sight target.

[0010] The beneficial effects of this invention are as follows: This invention provides a non-line-of-sight (Line-of-Sight) moving target 3D reconstruction method based on MIMO millimeter-wave radar, which can perform unambiguous estimation of DoD / DoA of multipath signals and achieve 3D reconstruction of non-line-of-sight targets. Specifically, a non-line-of-sight multipath propagation model is first established, and the multipath signal model of MIMO millimeter-wave radar is derived. Furthermore, an unambiguous estimation algorithm for DoD / DoA of multipath signals is proposed, solving the problem of distinguishing between the main lobe and grating lobe in the DoD / DoA estimation process of existing commercial MIMO radars. Finally, a path-guided non-line-of-sight target 3D localization algorithm is proposed, which uses different paths to locate non-line-of-sight targets separately, and fuses the localization results of different paths to form a 3D point cloud. The method of this invention has the following advantages:

[0011] 1. This invention can be used for unambiguous DoD / DoA estimation of multipath signals in co-located MIMO radar;

[0012] 2. This invention can use multiple paths to locate non-line-of-sight targets, so it remains effective even when a single path is missing.

[0013] 3. This invention can achieve three-dimensional reconstruction of non-line-of-sight targets by fusing multiple paths, forming a dense point cloud.

[0014] 4. This invention can be applied to fields such as driver assistance. Attached Figure Description

[0015] Figure 1 The processing flow of the proposed method is described below.

[0016] Figure 2 This is a schematic diagram of non-line-of-sight multipath signal propagation.

[0017] Figure 3 This is a schematic diagram of a MIMO millimeter-wave radar array antenna.

[0018] Figure 4 This is a schematic diagram of a fuzzy-free estimation algorithm for DoD / DoA of multipath signals.

[0019] in, Figure 4 (a)(c)(e) are fuzzy two-dimensional DoD / DoA spectra. Figure 4 (b)(d)(f) are the estimation results obtained using the method of the present invention.

[0020] Figure 5 To utilize the path and path A schematic diagram of three-dimensional positioning of a non-line-of-sight target.

[0021] in, Figure 5 (a) and Figure 5 (b) are schematic diagrams of three-dimensional positioning of non-line-of-sight targets using distance, angle measurement and angle measurement only, respectively.

[0022] Figure 6 This is the experimental scenario.

[0023] in, Figure 6 (a) is a photo of the experimental scene. Figure 6 (b) is a schematic diagram of the experimental scenario.

[0024] Figure 7 This is an echo signal from a non-line-of-sight target.

[0025] in, Figure 7 (a) is the distance-Doppler plot. Figure 7 (b) is a multi-period distance image.

[0026] Figure 8 The result is a 3D reconstruction of a non-line-of-sight target.

[0027] in, Figure 8 (a) shows the reconstruction result from viewpoint 1. Figure 8 (b) shows the reconstruction result from viewpoint 2.

[0028] Figure 9 It is the empirical cumulative distribution function (ECDF) for non-line-of-sight target positioning errors. Detailed Implementation

[0029] To facilitate understanding of the technical content of this invention by those skilled in the art, the following description, in conjunction with the accompanying drawings, further illustrates the invention.

[0030] like Figure 1 As shown, the present invention provides a non-line-of-sight moving target 3D reconstruction method based on MIMO millimeter-wave radar, comprising the following steps:

[0031] Step 1: Non-line-of-sight multipath propagation model

[0032] Consider as Figure 2 The scenario shown includes an L-shaped wall formed by the first and second walls, and a third wall parallel to the second wall. All three walls have surfaces perpendicular to the ground. The radar is positioned directly in front of the first wall. Therefore, the second wall is a non-line-of-sight wall, and the third wall is a line-of-sight wall. (Moving target) Located between the line-of-sight wall and the non-line-of-sight wall. Due to the presence of the building, the radar... With the goal The line-of-sight path between the electromagnetic waves is blocked, causing them to reach the target only through multiple reflections on the wall. Considering that millimeter-wave signals often experience significant power attenuation during reflection, this invention only considers the following two one-way propagation paths:

[0033] (1) Path Electromagnetic waves reach the target after being reflected once by a wall at line of sight.

[0034] (2) Path Electromagnetic waves are reflected once by a line-of-sight wall and then reach a non-line-of-sight wall, after which they are reflected again by the non-line-of-sight wall and reach the target.

[0035] Assuming the viewing distance and reflective surface parameters are known. Non-line-of-sight reflective surface parameters ,in Let represent the perpendicular distance between the origin of the coordinate system and the reflecting surface, and the normal vector of the reflecting surface, respectively. The Householder matrix is ​​defined as follows:

[0036]

[0037] in, Represents the identity matrix. For the normal vector of the reflecting surface and The angle between the positive axes. The reflecting surface here should be understood as either a line-of-sight reflecting surface or a non-line-of-sight reflecting surface. When it is a line-of-sight reflecting surface, the parameter subscript is LOS, and when it is a non-line-of-sight reflecting surface, the parameter subscript is NLOS.

[0038] Combined with radar position Using Householder transformation, the path corresponding to the transformation is derived. and Virtual radar location and

[0039]

[0040] It is evident that the multipath effect effectively provides radar with more virtual perspectives for target observation. Unlike bistatic and multistatic radar receivers, which can separate signals from different transmitting nodes, multipath echoes from non-line-of-sight targets coexist in the received signal of a single radar, requiring further sorting and identification to achieve the localization of non-line-of-sight targets.

[0041] Non-line-of-sight target localization requires the use of geometric parameters, including multipath distance and azimuth information. The path of electromagnetic waves must also be considered. Upon reaching the target, after being scattered by the target, it travels along the path Returning to radar, among which For ease of description, this invention will denote the two-way path as... The corresponding pseudo-bistatic radius (PBR) is given by the following formula.

[0042]

[0043] in, Describing the l2 norm, and Representing paths and path The length of the target. Furthermore, the direction vector between the non-line-of-sight target and the virtual radar can be expressed as...

[0044]

[0045] Step 2: MIMO millimeter-wave radar multipath signal model

[0046] Considering a single MIMO millimeter-wave radar, its antenna array is as follows: Figure 3 As shown, the radar transmitting array The array elements are distributed in At different pitch heights, its first The number of launch elements at each elevation altitude is The receiving array is A linear array of elements. The center frequency of the transmitting antenna is... bandwidth is The frequency modulation period is The linear frequency modulated continuous wave (LFMCW) signal is generated by the first... Signals emitted by the high-altitude transmission array are scattered by non-line-of-sight targets and return to the radar. Path echo It can be represented as

[0047]

[0048] in , and Represent the scattering coefficients and path of non-line-of-sight targets, respectively. Reflection attenuation and path Reflection attenuation; , Separate paths azimuth and elevation angles Then the path azimuth angle, Indicates the first The elevation baseline length of each transmission array; Indicates the first A height-controlled transmission array guide vector, Indicates the receiving guide vector; The radial velocity of the non-line-of-sight target relative to the radar; and These represent the sampling interval and the pulse repetition time (PRT), respectively. Indicates the frequency modulation slope. λ represents the speed of electromagnetic wave propagation, and λ represents the wavelength of the transmitted signal. It represents the Kronecker product.

[0049] Assuming the scenario contains If there are non-line-of-sight targets, the received signal can be expressed as:

[0050]

[0051] in Indicates the first The target comes from the path The echo, and These represent static environmental echo and Gaussian white noise, respectively. Considering the particularly weak signal-to-noise ratio (SNR) of non-line-of-sight target echoes, range-Doppler processing is first applied to the received signal. Beat frequency is also considered. Doppler frequency , , , and Let these represent the number of beat frequency points and the number of Doppler frequency points, respectively. The corresponding matched filter output can be expressed as:

[0052]

[0053] in and These represent the number of grids for the intermediate frequency and Doppler frequency under consideration, respectively.

[0054] Based on the range-Doppler processing results, a constant false alarm rate (CFAR) detector is used to extract potential non-line-of-sight moving targets. At the time of its establishment, it was believed that the first There is a target in each of the range-Doppler cells, among which The CFAR detection threshold is used. Based on the detection results, the angle parameters of each target are further estimated to achieve the localization of non-line-of-sight targets. The CFAR detection threshold is a known existing technology. In practical applications, the corresponding CFAR detection threshold is set according to the considered false alarm probability (Pfa). This invention will not provide a detailed explanation of the CFAR detection threshold value.

[0055] Step 3: Unambiguous estimation of DoD / DoA for multipath signals

[0056] like Figure 4 As shown, unlike the traditional radar target angle estimation problem, multipath signal angle estimation requires additional consideration of the following issues:

[0057] (1) Two-dimensional angle estimation: The departure path and arrival path of a multipath signal may be different, making the departure angle (DoD) and arrival angle (DoA) unequal;

[0058] (2) Unambiguous angle estimation: The element spacing of most commercial MIMO radar transmitting arrays is often several times half a wavelength, resulting in periodic spectral peaks in the DoD dimension, making it difficult to distinguish the main lobe from the grating lobe.

[0059] To address the aforementioned problems, this invention proposes a fuzz-free DoD / DoA estimation algorithm for multipath signals. Specifically, the DoD / DoA estimation problem for multipath signals can be described as follows:

[0060]

[0061] in Indicates the complex amplitude of the received signal. , Represents the Kronecker product. and Let the transmit steering matrix and receive steering matrix be represented respectively, and defined as follows:

[0062]

[0063] in and These are the sets of departure angles and arrival angles, respectively. The number of angles to be considered. In practical applications, for omnidirectional antennas, the range is from -90° to +90°. If each angle is 1°, then Na = 181.

[0064] according to Figure 4 The DoD / DoA fuzzy estimation results for the multipath signal shown indicate that, in the two-dimensional DoD / DoA spectrum, the main lobe is located on the diagonal of the spectrum or is symmetrical about it, with the main lobe located on the diagonal being a special case of the latter. Based on this characteristic, the correlation between all candidate DoD / DoA pairs and the residuals of the array signal is first calculated, which can be expressed as follows:

[0065]

[0066] in, Candidates for DoD / DoA , ; They are respectively The estimation result in the t-th iteration; For transmit-receive steering vector, Indicates the first The array signal residual at the next iteration, its initial value The superscript H indicates conjugate transpose.

[0067] 1) When At that time, based on the characteristic that the main lobe is symmetric about the diagonal of the spectrum, the estimation result can be uniquely determined, and the estimation result of the current iteration is taken as... , ,in and Let each represent the estimated set of angles, with initial values ​​of . , This means that the initial value is empty.

[0068] 2) When At this time, there are two cases: either the grating lobe is located on the diagonal or the main lobe is located on the diagonal, as shown below. Figure 4 (c) and Figure 4 As shown in (e). For specific differentiation, based on the periodicity of the DoD dimension spectral peaks, it is assumed that... From the DoD corresponding to the main lobe, we can deduce the position of its raster lobe.

[0069]

[0070] in, This indicates the element spacing of the transmitting antenna array. This indicates that the angle is less than 0, suggesting that the grating lobe should be located to its right. The angle is greater than 0, suggesting that the grating lobe should be located to its left.

[0071] Furthermore, the following criteria are used to determine whether the hypothesis holds true.

[0072]

[0073] in In this embodiment, the threshold is determined. The value is 15dB; when At the time of its establishment, The DoD corresponding to the main lobe is updated in the current iteration as follows: , ;when At the time of its establishment, For the DoD corresponding to the grating lobe, the DoD of its main lobe should be... Therefore, the estimation result is updated to , . This should be understood as Assumption 1, that is... Main petal; This should be understood as assumption 0, that is For grating lobes.

[0074] After obtaining the estimates of DoD and DoA, update the complex amplitude. and residual

[0075]

[0076] in This indicates the search for a pseudo-inverse.

[0077] go through After the first iteration, the iteration is terminated, and the angle estimation result is obtained. , as well as , Represents the set of real numbers. Let represent the set of complex numbers, where and The elements in the table correspond one-to-one, forming DoD / DoA angle pairs. In practical applications... The number of signals is determined based on the estimated number of signals; for example, if there is one target in the signal, then... Set to 1; set to 2 for 2 targets.

[0078] For the One DoD / DoA angle pair Its pitch angle It can be obtained through the following formula

[0079]

[0080] The superscript H indicates conjugate transpose. , Indicates the pitch steering vector. For the first The phase difference of each transmitting antenna array relative to the reference position can be expressed as: .

[0081] PBR estimation: For That is, the data of the z-th Doppler cell and the k-th range cell of the m-th height array, whose corresponding PBR can be calculated as follows: , This represents the estimated PBR.

[0082] Step 4: Path-guided non-line-of-sight target 3D localization algorithm

[0083] The estimated PBR, azimuth DoD, elevation DoD, and azimuth DoA are respectively denoted as... , , and Combining Figure 5 This further illustrates the 3D localization algorithm for path-guided non-line-of-sight targets.

[0084] (1) Path-based Non-line-of-sight target 3D localization: If The path is identified as a path ,therefore , , Based on distance and angle measurements, the position of the non-line-of-sight target is derived.

[0085]

[0086] in Represents direction vector The estimate is derived from the following formula.

[0087]

[0088] in Representing a path Relative to the line-of-sight direction of the radar.

[0089] (2) Path-based Non-line-of-sight target 3D localization: If The path is identified as a path ,therefore , , Because non-line-of-sight targets are not only located in areas determined by the virtual radar position... , and pseudo-bibase distance On the determined ellipsoid, it is also located by and leaving the corner On the determined ray, such as Figure 5 As shown in (a), the position of a non-line-of-sight target can therefore be determined by solving for the intersection of the ellipsoid and the ray. This invention proposes to utilize only the estimated angle... , and Achieve localization of non-line-of-sight targets without distance measurement. For example... Figure 5 As shown in (b), the angle of arrival of multipath signals is considered. Non-line-of-sight targets will be located at... The vertex is [a], and the cone angle is [b]. The ray intersects the conical surface, thus the non-line-of-sight target can be located by solving the following system of equations.

[0090]

[0091] in This represents the direction vector of the cone axis.

[0092] (3) Path-based Non-line-of-sight target 3D localization: If The path is identified as a path ,therefore , , . and path-based Similar to the three-dimensional localization of non-line-of-sight targets, the position of non-line-of-sight targets can be obtained by solving the following system of equations.

[0093]

[0094] in Indicates As the vertex, The cone's axial direction vector is the cone angle. Represents direction vector The estimate, , .

[0095] Based on the above method, the detected range-Doppler cells are processed one by one to obtain the paths. , and The corresponding point clouds are then merged to achieve 3D reconstruction of non-line-of-sight targets.

[0096] The following is a specific implementation of the present invention based on an experimental example.

[0097] Experimental scenarios such as Figure 6 As shown, the sensor used in the experiment was a TI AWR2243 Cascade radar, placed at (0.3 m, 0.5 m), with the array making a 55° angle with the negative x-axis. The x-coordinates of the line-of-sight wall and the non-line-of-sight wall were 1.2 m and 7.96 m, respectively, with corresponding polar coordinate parameters of... and The two moving targets are located at... and During the experiment, the radar remained stationary, and the transmitting antenna emitted an LFMCW signal with the following waveform parameters: initial frequency 77 GHz, bandwidth 2.05 GHz, sampling frequency 20 MHz, number of sampling points 512, pulse repetition period 420 μs, and number of pulses 128.

[0098] Figure 7 The echo signal of the non-line-of-sight target is given. Figure 7 (a) Distance-Doppler plot, Figure 7 (b) is a multi-period range map. Non-line-of-sight (NLS) targets are detected and extracted from the range-Doppler map, and their DoD and DoA are estimated to further derive their locations. A point cloud of NLS targets is formed through point-by-point processing, such as... Figure 8 As shown, the reconstructed result is similar to the actual target shape. The target point cloud is projected onto a two-dimensional plane, and the two-dimensional positioning error of this invention for non-line-of-sight targets is statistically analyzed. Figure 9 The corresponding empirical cumulative distribution function (ECDF) for positioning error is given, showing that path-based... The location result is better than the path The reason may be the path The echo has a higher signal-to-noise ratio, making the corresponding measurements more accurate. Furthermore, the proposed invention utilizes a path... and path The root mean square error (RMSE) for the positioning of target 1 is 0.078 m and 0.275 m, respectively, while the RMSE for the positioning of target 2 is 0.148 m and 0.368 m, respectively.

[0099] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of the claims of the invention.

Claims

1. A method for non-line-of-sight (NLOS) moving target three-dimensional reconstruction based on MIMO millimeter wave radar, characterized in that, Comprise: S1, construct a non-line-of-sight multipath propagation model, and derive a multipath echo signal model of the MIMO millimeter wave radar based on the constructed non-line-of-sight multipath propagation model; The non-line-of-sight multipath propagation model is based on a building model, which comprises a first wall and a second wall forming an L-shaped wall, and a third wall parallel to the second wall, surfaces of the three walls are perpendicular to the ground, and a radar is arranged in front of the first wall, so that the second wall is a non-line-of-sight wall, and the third wall is a line-of-sight wall; a moving target is located between the line-of-sight wall and the non-line-of-sight wall, and comprises two one-way propagation paths: path : electromagnetic waves are reflected once on the line-of-sight wall to reach the target; and path : electromagnetic waves are reflected once on the line-of-sight wall and then reach the non-line-of-sight wall, and then are reflected on the non-line-of-sight wall to reach the target. S2, performing range-doppler processing on the received signal represented by the multipath echo signal model in step S1, and detecting a non-line-of-sight moving target; S3, according to the detection result of step S2, performing unambiguous estimation on the DoD / DoA pair of the multipath signal; S4, DoD / DoA estimated based on S3, non-line-of-sight multipath propagation model established in combination with S1, respectively using paths , and to locate the target, and fuse the point clouds of the three to realize three-dimensional reconstruction of the non-line-of-sight target, represents that the electromagnetic wave reaches the target through path , and then is scattered by the target to return to the radar through path ; represents that the electromagnetic wave reaches the target through path , and then is scattered by the target to return to the radar through path , represents represents that the electromagnetic wave reaches the target through path , and then is scattered by the target to return to the radar through path .

2. The method of claim 1, wherein, The array elements of the radar transmitting array are distributed at multiple different elevation heights, and the receiving array is a linear array of a plurality of array elements.

3. The method of claim 2, wherein, The multipath echo signal model in step S1 is represented as: ； where denotes the electromagnetic wave propagation path to the target, scattered by the target, and propagating back to the radar, , and denote the scattering coefficient of the non-line-of-sight target, the reflection attenuation of the path , and the reflection attenuation of the path , respectively; , denote the azimuth and elevation angles of the path , denote the azimuth angle of the path , denotes the elevation baseline length of the th transmit array; denotes the elevation transmit array steering vector of the th transmit array, denotes the receive steering vector; is the radial velocity of the non-line-of-sight target relative to the radar; and denote the sampling interval and the pulse repetition period, respectively, denotes the frequency modulation slope, denotes the electromagnetic wave propagation speed, and λ denotes the transmitted signal wavelength, denotes the Kronecker product.​ 4. The method of claim 3, wherein, A constant false alarm rate detector is used to extract potential non-line-of-sight moving targets.

5. The method of claim 4, wherein, Step S3 includes the following steps: S31, the DoD / DoA pair estimation problem of the multipath signal is represented as: ； wherein denotes a complex amplitude of a received signal, denotes a range-Doppler processed received signal, denotes an l2 norm, , denotes a Kronecker product, and denote a transmit steering matrix and a receive steering matrix, respectively, and are sets of DoD and DoA, respectively; S32, based on the characteristics that the main lobe is located on the diagonal line of the two-dimensional DoD / DoA spectrum or is symmetric about it, the correlation degree of all candidate DoD / DoA pairs with the array signal residual is calculated, which is represented as: ； wherein is the transmit-receive steering vector, and the superscript H denotes the conjugate transpose, denotes the array signal residual at the th iteration, which is initialized as ; S33、when the main lobe is located on the diagonal of the spectrum or symmetric about it, the estimation result is uniquely determined, and the estimation result of the next iteration is taken as , where and respectively represent the estimated DoD and DoA sets of the current iteration; S34、when there are two cases, the grid lobe is on the diagonal or the main lobe is on the diagonal, the DoD corresponding to the main lobe, when the estimation result of the secondary iteration is updated as , ; the DoD corresponding to the grid lobe, the DoD of the main lobe should be , so the estimation result is updated as , ; S35. Update the complex amplitudes based on the estimation results of the DoD and DoA of the current iteration and the residual ; S36, according to the estimation results of DoD and DoA obtained by the final iteration and the complex amplitude, the elevation angle of each DoD / DoA pair is calculated; S37, the estimated pseudo-baseline distance is derived according to the beat frequency.

6. The method of claim 5, wherein, Step S4 specifically includes the following steps: S41, based on the path of the non-line-of-sight target: if , the path is identified as the path , thus , , ; from the distance and angle measurements, the position of the non-line-of-sight target is derived: ； wherein represents a direction vector estimate of the direction S42, based on the path of the non-line-of-sight target: if , the path is identified as the path , thus , , ; considering the angle of arrival of the multipath signal , the non-line-of-sight target will be located on a conical surface with as the apex and the conical angle , thus solving the intersection of the ray and the conical surface can locate the non-line-of-sight target; S43, based on the path of the non-line-of-sight target: if , the path is identified as the path , thus , , ; considering the angle of arrival of the multipath signal , the non-line-of-sight target will be located on a conical surface with as the apex and the conical angle , thus solving the intersection of the ray and the conical surface can locate the non-line-of-sight target; S44, based on steps S41-S43, the detected distance-Doppler units are processed one by one, and the paths , and corresponding point clouds are obtained respectively, and finally the point clouds of the three are fused to realize three-dimensional reconstruction of the non-line-of-sight target.

Images