Method for tracking unknown multi-target motion trail in non-line-of-sight scene

Abstract

The invention discloses a method for tracking an unknown multi-target motion track in a non-line-of-sight scene, and the method comprises the steps: collecting a target echo signal in a non-line-of-sight region through employing a frequency modulation continuous wave radar, carrying out the preprocessing of the echo signal, and obtaining a time distance map containing the motion features of a human body target; an improved self-adaptive trace extraction algorithm based on gridding and conditional constraint is adopted to extract a target motion trace; by adopting the fusiform front trace energy suppression algorithm provided by the invention, cross interference among different traces is reduced; according to the Doppler consistency in the trace segments and the Doppler similarity between the traces, trace attribution judgment and target attribution judgment are completed, and meanwhile target counting is completed; determining each multi-path echo order by adopting a trace order matching algorithm; and an elliptical cross positioning algorithm is adopted, and a multipath propagation model is combined to complete target positioning and reconstruct a target motion trail. The method is suitable for tracking unknown multi-target motion tracks in a non-line-of-sight scene, has the advantages of no need of target prior information, high robustness, high precision and the like, and can be widely applied to the fields of automatic driving, security monitoring, disaster rescue and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal processing and target tracking, and in particular to a method for tracking motion trajectories of unknown multiple targets in a non-line-of-sight scenario. Background Art

[0002] With the increasing demand for concealed target perception capabilities in areas such as autonomous driving, security monitoring, and disaster relief, human target detection and tracking in non-line-of-sight environments has become a research hotspot. Frequency-modulated continuous wave radar (FMCW) is widely used in these tasks due to its ability to penetrate obstacles and operate 24 / 7. Compared to traditional optical or laser sensors, FMCW radar can stably detect targets in complex environments and has inherent advantages in privacy protection.

[0003] Existing radar-based non-line-of-sight (NLOS) tracking methods often rely on fully extracted multipath information and typically assume a known number of targets or simple motion. However, in practical applications, NLOS scenarios present challenges such as an unknown number of targets, severe cross-track interference, and difficulty extracting tracks. These issues lead to significant errors in target attribution and trajectory reconstruction using traditional methods.

[0004] Therefore, in order to address the problems of unknown number of targets, complex multipath environment, and difficulty in extracting trajectories in non-line-of-sight scenarios, it is urgent to propose a radar signal processing method that can efficiently and stably track the motion trajectories of unknown multiple targets in non-line-of-sight environments. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for tracking the motion trajectories of unknown multiple targets based on frequency modulated continuous wave radar in a non-line-of-sight scenario to address the problems existing in the prior art, such as the unknown number of targets, difficulty in trajectory extraction, and difficulty in target attribution due to trace intersection.

[0006] The technical solution to achieve the purpose of the present invention is: a method for tracking the motion trajectories of unknown multiple targets in a non-line-of-sight scenario, the method comprising the following steps:

[0007] Step 1: Use a frequency modulated continuous wave radar to collect target echo signals in a non-line-of-sight area, and preprocess the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing human target motion characteristics;

[0008] Step 2: Extract the target motion trace using an improved adaptive trace extraction algorithm based on gridding and conditional constraints. The gridding algorithm divides the time axis and distance axis into unit grids based on human body size and motion characteristics. The conditional constraints are used to deal with trace jumps and trace breaks, including acceleration condition constraints and trace break tolerance constraints.

[0009] Step 3: Using the spindle-shaped front trace energy suppression algorithm to reduce cross-interference between different traces, repeat steps 2 and 3 until all traces are extracted;

[0010] Step 4: Based on the Doppler consistency within the trace segment and the Doppler similarity between traces, the trace segment ownership determination and target ownership determination are completed, and the target counting is completed at the same time;

[0011] Step 5: Determine the order of each multipath echo using a trace order matching algorithm;

[0012] Step 6: Target positioning is achieved and the target motion trajectory is reconstructed through the ellipse intersection positioning algorithm combined with the multipath propagation model and trace order information.

[0013] Furthermore, step 1 specifically includes:

[0014] Step 1-1: Use a frequency-modulated continuous wave radar to collect target echo signals in the non-line-of-sight area and rearrange the echo signals into an N×M matrix s(m,n). The vertical dimension of the rearranged data matrix is ​​the fast time dimension, and the horizontal dimension is the slow time dimension. The fast time dimension is divided according to the pulse sampling points, and the slow time dimension is divided according to the inter-pulse transmission interval. n represents the fast time sampling point index, and m represents the slow time transmission pulse index.

[0015] Step 1-2, preprocessing the echo signal matrix, including DC component suppression and moving target detection;

[0016] Steps 1-3: pre-process the echo matrix Perform fast Fourier transform to obtain the time distance graph matrix The calculation formula is:

[0017]

[0018] Where t is the time unit index, r is the distance unit index, j is the imaginary unit, R and T are the distance index range and time index range in the real number field, respectively.

[0019] Furthermore, the DC component suppression in step 1-2 adopts the mean filtering method, that is, the row mean is calculated along the slow time dimension of the echo signal matrix, and the row mean is subtracted from the original data to obtain the echo matrix after the DC component is suppressed, specifically:

[0020]

[0021] Wherein, M is the slow time transmission pulse index.

[0022] Furthermore, the moving target detection in step 1-2 is specifically as follows: the echo matrix after DC component suppression is By offsetting and subtracting along the slow time axis, the stationary target echoes with the same amplitude and phase are eliminated, and the moving targets with continuously changing phases are retained to obtain the echo matrix after moving target detection. The calculation formula is as follows:

[0023]

[0024] Where n = [1, ..., N] represents the fast time sampling point, and m = [1, 2, ..., M] represents the slow time transmission pulse index.

[0025] Furthermore, step 2 specifically includes:

[0026] Step 2-1: On the time-distance diagram, grid the time and distance axes according to the spatial size of the human target. The time axis is divided into fixed time windows of length τ, where the value of τ is related to the inherent characteristics of human motion; the distance axis is divided into fixed distance intervals h, where h is related to human size. Gridding is performed along the fast time dimension of the time-distance diagram, using a grid size of τ × h and setting the grid sampling overlap based on the energy background information.

[0027] Step 2-2: Based on the gridded time-distance graph, calculate the energy accumulation value of each grid in each time column to obtain the energy curve E corresponding to the current time column. n , and its calculation formula is:

[0028]

[0029] Among them, n' is the n'th grid index in the current time column, t represents the current indexed time column, i.e., the time unit index; N' is the total number of grids;

[0030] Step 2-3: In each time window, the peak search method is used to find the energy curve E n Select the local maximum as the candidate grid r ij , specifically:

[0031] R n ={r 1j ,r 2j ,…,r ij}=findpeaks(E ij ,E th ,h)

[0032] Where i represents the i-th time column of the time distance graph, j represents the j-th grid of the i-th time column, and E ij is the current grid r ij The energy value, E th is the grid screening threshold, h is the human body size, and the selected grid r ij Store in grid matrix Rn ;

[0033] Steps 2-4: Apply conditional constraints on the candidate grid, specifically including:

[0034] Step 2-4-1, introduce acceleration constraints, perform second-order differences on three consecutive grids along the time direction, and calculate the current grid r ij The trajectory acceleration a ij , whose expression is:

[0035]

[0036] Among them, τ' is the time interval between adjacent grids in the time dimension, are the extracted grids on the time columns (j-1)τ' and (j-2)τ' respectively;

[0037] Step 2-4-2, apply energy penalty factor to the corresponding candidate trace point The updated grid energy is:

[0038]

[0039] Where λ is the acceleration penalty coefficient, which is used to control the smoothness of the trace, G is the updated grid matrix, and E[·] represents the energy value in the grid; r j is the jth time column of the current energy graph E, k represents the kth grid of the jth time column, E[r j ,k] represents the energy value of the kth grid on the jth time column;

[0040] Step 2-4-3, introduce the trace break tolerance constraint to deal with the trace break situation. When the break length n of the trace segment g Greater than the trace break tolerance threshold n th , delete the current trace;

[0041] Step 2-5: On the updated grid matrix G, the trace finding problem is transformed into an optimization problem, and a set of grid points that maximizes the trace energy is extracted as the trace extraction result. Its expression is:

[0042]

[0043] Where N is the number of grids in the current j-th time column.

[0044] Furthermore, step 3 specifically includes:

[0045] Step 3-1: Starting from the center of the extracted trace grid point, a spindle-shaped energy suppression region is constructed. The suppression region is symmetrically extended in the time direction, and the energy in the grid decreases from the horizontal center line of the grid to both sides.

[0046] Step 3-2: In the spindle-shaped inhibition region, the energy distribution of the original time-distance map Z(t,r) is corrected. The corrected energy Z * (t,d) is calculated according to the following formula:

[0047]

[0048] Among them, t and d represent the time column index and the grid index under the current time column in the time distance graph respectively. represents the extracted grid in time column j, and J0(·) is the zero-order Bessel function, specifically:

[0049]

[0050] Among them, x is used to represent variables and has no other meaning;

[0051] Step 3-3, repeating steps 3-1 to 3-2, and performing spindle energy suppression processing on the extracted traces in sequence;

[0052] Step 3-4: Repeat step 2 and step 3-1 to step 3-3 to extract all traces in the time-distance graph.

[0053] Furthermore, step 4 specifically includes:

[0054] Step 4-1: Extract the envelope curve of the short-time Fourier transform of the echo trace and perform a fast Fourier transform on the envelope curve to obtain the Doppler characteristic value of each echo trace. If there is any intersection of the traces, determine the trace segment attribution according to the following steps;

[0055] Step 4-1-1, assume that there are q crossing trace segments near the intersection, with start and end times m and m respectively. s1 ,m s2 ,…,m sq With m e1 ,m e2 ,…,m eq , extract the envelope curve of the short-time Fourier transform of these q cross trace segments in their respective corresponding time intervals, and perform fast Fourier transform on the envelope curve to obtain the Doppler eigenvalue sequence of q echo trace segments, denoted as f d1 ,…,f dq ;m sq 、m eq are the start and end times of the qth cross trace segment, f dq is the Doppler eigenvalue of the qth cross-trace segment;

[0056] Step 4-1-2, determine whether the cross trace segments belong to the same trace based on the Doppler eigenvalue. When |f di -f dj When |<ε, the cross-trace segments i and j are considered to belong to the same trace, which is the Doppler consistency within the trace segment; where f di 、f dj are the Doppler characteristic values ​​of the i-th and j-th cross trace segments, respectively, and ε is the Doppler characteristic judgment threshold;

[0057] Step 4-2: determine whether different traces belong to the same target based on the Doppler eigenvalues ​​between the traces. This is the Doppler similarity between the traces.

[0058] In step 4-3, the trace sets belonging to the same target are counted, and the number of trace groups obtained is the target counting result.

[0059] Furthermore, step 5 specifically includes:

[0060] Step 5-1: For each trace that has been determined to belong to a target, record the energy sum of each trace corresponding to each target in the time interval. Assuming that there are s traces corresponding to target b, calculate the energy sum of these s traces to obtain the energy sequence E b1 ,…,E bs , sort the energy sequence in descending order; according to the different echo reflection orders, classify the echo signals according to the reflection order and obtain the order matrix Ψ:

[0061]

[0062] Among them, l ij Indicates that the number of reflections of the transmitted signal is i and the number of reflections of the reflected signal is j, i, j = 1, 2, 3;

[0063] Traverse all i,j values ​​and construct a multipath set consisting of different reflection times;

[0064] Step 5-2: In the energy sequence sorted in descending order, b1 As l 11 , with E b2 As l 12 , solve the target position through the ellipse cross positioning algorithm and obtain the virtual target point T * , according to the virtual target point position and the radar position, combined with the multipath propagation model, the propagation distance of each order multipath corresponding to the virtual target point is obtained, and the difference d between the propagation distance corresponding to the virtual target point and the propagation distance obtained by the multipath propagation model is calculated;

[0065] Step 5-3: traverse all multipath sets, repeat step 5-2, and select the set with the smallest difference d as the trace order matching result;

[0066] In step 5-4, the order information of each trace corresponding to each target is recorded as necessary information for subsequent target positioning and trajectory reconstruction.

[0067] Furthermore, step 6 specifically includes:

[0068] Step 6-1: Group the extracted traces according to their reflection order information, where the first-order reflection trace corresponds to the single-reflection echo, and the first- and second-order mixed reflection trace corresponds to the echo that first reaches the target after a single reflection and then reaches the radar sensor after a second reflection;

[0069] Step 6-2: For the first-order reflection trace, calculate the radial distance d1 from the target to the radar based on the echo propagation time t1 and propagation velocity c1:

[0070]

[0071] Step 6-3: For the first-order and second-order mixed reflection traces, an elliptical structure is formed between the target and the two reflecting walls. The total one-way propagation distance d of the first-order trace and the first-order and second-order mixed traces is * for:

[0072] d * =d r11 +d r12

[0073] Among them, d r11 with d r12 They represent the one-way distance of the first-order reflection trace and the one-way distance of the first- and second-order mixed reflection trace, d * is the major axis of the ellipse;

[0074] Step 6-4: Use d1 obtained in step 6-2 as the radius of the circle and the virtual radar coordinate point as the center of the circle to construct circle C1. * As the major axis of the ellipse, the first virtual radar coordinate point and the second virtual radar coordinate point are used as the two foci of the ellipse to construct the ellipse E1;

[0075] In step 6-5, the intersection of the first-order circle equation and the second-order ellipse equation obtained in step 6-4 is solved as the estimated position of the target in the current time unit. The estimated positions of the target in all time units are further connected in series to form a continuous motion trajectory of the target.

[0076] In another aspect, a system for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario is provided, the system comprising sequentially executing:

[0077] The first module is used to use a frequency modulated continuous wave radar to collect target echo signals in the non-line-of-sight area and pre-process the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing human target motion characteristics;

[0078] The second module is used to extract the target motion trajectory using an improved adaptive trajectory extraction algorithm based on gridding and conditional constraints. The gridding is to divide the time axis and distance axis into unit grids based on human body size and human motion characteristics. The conditional constraints are used to deal with trajectory jumps and trajectory breaks, including acceleration condition constraints and trajectory break tolerance constraints.

[0079] The third module is used to reduce the cross interference between different traces through the spindle-shaped front trace energy suppression algorithm, and repeatedly execute the second and third modules until all traces are extracted;

[0080] The fourth module is used to complete the track segment ownership determination and target ownership determination based on the Doppler consistency within the track segment and the Doppler similarity between tracks, and simultaneously complete the target counting;

[0081] The fifth module is used to determine the order of each multipath echo through a trace order matching algorithm;

[0082] The sixth module is used to realize target positioning and reconstruct the target motion trajectory through the elliptical cross positioning algorithm combined with the multipath propagation model and trace order information.

[0083] Compared with the prior art, the present invention has the following significant advantages:

[0084] 1) This paper proposes an improved trace extraction algorithm based on gridding and conditional constraints. Based on the traditional trace extraction algorithm, gridding, acceleration constraints and trace breakage tolerance constraints are proposed in combination with the inherent characteristics of human targets. This algorithm can achieve stable extraction of unknown multi-target traces in non-line-of-sight environments. Compared with the traditional trace extraction method based on energy detection, it significantly improves the accuracy of trace extraction and can accurately extract target traces in complex scenarios such as trace crossing and trace breakage.

[0085] 2) The present invention introduces the spindle-shaped front trace energy suppression mechanism for the first time in the trace extraction process, which can effectively reduce the energy interference in the trace intersection area and improve the accuracy of unknown multi-target separation and attribution determination.

[0086] 3) The present invention completes the trace attribution determination and target attribution determination based on the Doppler consistency within the trace segment and the Doppler similarity between traces, and simultaneously completes the target counting. It can accurately complete the trace attribution and target counting tasks when the number of targets is unknown.

[0087] 4) The present invention innovatively introduces the elliptical cross-location (ECL) method into the trajectory order inference. By jointly solving the first-order trajectory and the second-order mixed trajectory, it can realize target positioning and motion trajectory reconstruction in a complex multipath propagation environment.

[0088] The present invention is described in further detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 The figure is a flowchart of a method for tracking unknown multi-target motion trajectories in a non-line-of-sight scenario according to an embodiment.

[0090] Figure 2 FIG. 1 is a diagram of an experimental scene in an embodiment.

[0091] Figure 3 Flowchart of an improved adaptive trace extraction algorithm based on gridding and conditional constraints in one embodiment.

[0092] Figure 4 This is a diagram of trace extraction results in an experimental scenario in an embodiment.

[0093] Figure 5 FIG. 4 is a diagram showing the Doppler feature extraction result of a trace segment in one embodiment.

[0094] Figure 6 : It is the envelope curve of the short-time Fourier transform in the time interval corresponding to the four trace segments in an embodiment, and its corresponding fast Fourier transform result.

[0095] Figure 7 Schematic diagram of an ellipse intersection positioning algorithm in one embodiment.

[0096] Figure 8 This is a diagram of the positioning results of the present invention in an experimental scenario in an embodiment. DETAILED DESCRIPTION

[0097] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0098] It should be noted that if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0099] In one embodiment, a method for tracking the motion trajectories of unknown multiple targets using a frequency modulated continuous wave (FMCW) radar in a non-line-of-sight (NLOS) scenario is provided. In this embodiment, there are two human targets located in a non-line-of-sight (NLOS) area. Target 1 performs reciprocating motion, while Target 2 performs stationary motion. The method includes the following steps:

[0100] Step 1, such as Figure 2 As shown in the figure, the FMCW radar is set up at a corner 0.5m away from the wall and at a height of 0.8m. It collects target echo signals in the non-line-of-sight area and performs pre-processing on the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing the motion characteristics of human targets. Specifically:

[0101] Step 1-1: Use a frequency-modulated continuous wave radar to collect target echo signals in the non-line-of-sight area and rearrange the echo signals into a 64×25000 matrix s(m,n). The rearranged data matrix has a fast time dimension vertically and a slow time dimension horizontally. The fast time dimension is divided according to the pulse sampling points, and the slow time dimension is divided according to the inter-pulse transmission interval. n represents the fast time sampling point index, and m represents the slow time transmission pulse index.

[0102] Step 1-2 is to pre-process the echo signal matrix, including DC component suppression and moving target detection. DC component suppression uses the mean filtering method, that is, the row mean is calculated along the slow time dimension of the echo signal matrix, and the row mean is subtracted from the original data to obtain the echo matrix after DC component suppression. Specifically:

[0103]

[0104] Moving target detection. The echo matrix after DC component suppression By offsetting and subtracting along the slow time axis, the stationary target echoes with the same amplitude and phase are eliminated, and the moving targets with continuously changing phases are retained to obtain the echo matrix after moving target detection. The calculation formula is as follows:

[0105]

[0106] Where n = [1,…,64] represents the fast time sampling point, and m = [1,2,…,25000] represents the slow time transmission pulse index;

[0107] Steps 1-3: The pre-processed echo matrix is ​​processed along the fast time axis. Perform fast Fourier transform, select Hamming window as window function, set the number of Fourier transform points to 1024, and obtain a time distance map with a size of 1024×25000 The calculation formula is:

[0108]

[0109] Where t is the time unit index, r is the distance unit index, j is the imaginary unit, R and T are the distance index range and time index range in the real number field, respectively.

[0110] Step 2: Use the improved adaptive trace extraction algorithm based on gridding and conditional constraints to extract the target echo trace. The algorithm flow is as follows: Figure 3 The gridding is to divide the time axis and distance axis into grids based on the human body size and human motion characteristics. The conditional constraints are used to deal with trace jumps and trace breaks, including acceleration condition constraints and trace break tolerance constraints.

[0111] Step 2-1: On the time-distance graph Z(t,r), grid the time and distance axes according to the spatial size of the human target. The time axis is divided into fixed time windows of length τ = 250ms, where the value of τ is related to the inherent characteristics of human motion; the distance axis is divided into fixed distance intervals h = 30, where h is related to the width of the human body. Gridding is performed along the fast time dimension of the time-distance graph using a grid size of 250 × 30 and a grid overlap of 50% based on the energy background information.

[0112] Step 2-2: Based on the gridded time-distance graph, calculate the energy accumulation value of each grid in each time column to obtain the energy curve E corresponding to the current time column. n , and its calculation formula is:

[0113]

[0114] Among them, n' is the n'th grid index in the current time column, t represents the current indexed time column, i.e., the time unit index; N' is the total number of grids;

[0115] Step 2-3: In each time window, the peak search method is used to find the energy curve E n Select the local maximum as the candidate grid rij , specifically:

[0116] R n ={r 1j ,r 2j ,…,r ij}=findpeaks(E ij ,E th ,h)

[0117] Where i represents the i-th time column of the time distance graph, j represents the j-th grid of the i-th time column, and E ij is the current grid r ij The energy value, E th is the grid screening threshold. In this scenario, E th = 12000, h is the human body size, set to h = 30, the selected grid r mn Store in grid matrix R n ;

[0118] Steps 2-4, conditional constraints are applied to the candidate grids, which can be divided into the following steps:

[0119] Step 2-4-1: To ensure the continuity and smoothness of the extracted trace, an acceleration constraint is introduced. The second-order difference of three consecutive grids along the time direction is performed to calculate the current grid r. ij The trajectory acceleration a ij , whose expression is:

[0120]

[0121] Among them, τ' is the time interval between adjacent grids in the time dimension, are the extracted grids on the time columns (j-1)τ' and (j-2)τ', τ = 250ms;

[0122] Step 2-4-2, apply energy penalty factor to the corresponding candidate trace point The updated grid energy is:

[0123]

[0124] Where λ is the acceleration penalty coefficient, which is used to control the smoothness of the trace and can be set to 1 by default. G is the updated grid matrix, and E[·] represents the energy value in the grid. j is the jth time column of the current energy graph E, k represents the kth grid of the jth time column, E[r j ,k] represents the energy value of the kth grid on the jth time column;

[0125] Step 2-4-3, in order to deal with the trace breakage situation, the trace breakage tolerance constraint is introduced. When the breakage length n of the trace segment g Greater than the trace break tolerance threshold n th When , the current trace is deleted. The trace break tolerance threshold in this experimental scenario is set to 500ms.

[0126] Step 2-5: On the updated grid matrix G, the trace finding problem is transformed into an optimization problem, and a set of grid points that maximizes the trace energy is extracted as the trace extraction result. Its expression is:

[0127]

[0128] In this embodiment, according to step 2, a total of 6 traces are extracted, such as Figure 4 As shown, they are P1, P2, P3, P4, P5, and P6 respectively.

[0129] Step 3: During the trace extraction process, the spindle trace energy suppression mechanism is applied to reduce the cross-interference between different traces, such as Figure 4 The enlarged area on the lower right is shown. Specifically:

[0130] Step 3-1: Starting from the center of the extracted trace grid point, a spindle-shaped energy suppression region is constructed. The suppression region is symmetrically extended in the time direction, and the energy in the grid decreases from the horizontal center line of the grid to both sides.

[0131] Step 3-2: In the spindle-shaped inhibition region, the energy distribution of the original time-distance map Z(t,r) is corrected. The corrected energy Z * (t,d) is calculated according to the following formula:

[0132]

[0133] Among them, t and d represent the time column index and the grid index under the current time column in the time distance graph respectively. represents the extracted grid in time column j, and J0(·) is the zero-order Bessel function, specifically:

[0134]

[0135] Among them, x is used to represent variables and has no other meaning;

[0136] Step 3-3, repeating steps 3-1 to 3-2, and performing spindle energy suppression processing on the extracted traces in turn to reduce the cross interference caused by the previous trace on the subsequent trace extraction, while retaining the local energy characteristics around the trace.

[0137] Step 3-4: Repeat step 2 and step 3-1 to step 3-3 to extract all traces in the time-distance graph.

[0138] Step 4: Based on the Doppler consistency within the trace segment and the Doppler similarity between traces, the trace attribution determination and target attribution determination are completed, and target counting is completed at the same time. Specifically:

[0139] Step 4-1: Extract the envelope curve of the short-time Fourier transform of the echo trace and perform a fast Fourier transform on the envelope curve to obtain the Doppler eigenvalue of each echo trace. If there is any intersection of the traces, determine the trace segment attribution according to the following steps;

[0140] Step 4-1-1, such as Figure 5 As shown, taking the trace intersection Q2 as an example, there are 4 cross trace segments near the intersection, namely S1, S2, S3, and S4, with start and end times in m s1 ,m s2 ,m s3 ,m s4 With m e1 ,m e2 ,m e3 ,m e4 , each pair of start and end time {m sn ,m en} form a set of time intervals, extract the envelope curve of the short-time Fourier transform in the time interval corresponding to these four trace segments, and perform fast Fourier transform on the envelope curve. The result is as follows Figure 6 As shown, Figure 6 The left side shows the short-time Fourier transform results of the four cross trace segments extracted near the intersection point Q2, and the right side shows the fast Fourier transform results of the left envelope. The Doppler eigenvalue sequence of the four echo trace segments is obtained, which is recorded as f d1 …f d4 In this embodiment, the Doppler characteristic value sequence is {1.50 Hz, 1.25 Hz, 1.25 Hz, 1.50 Hz};

[0141] Step 4-1-2, determine whether the trace segments belong to the same trace based on the Doppler eigenvalue. When |f di -f dj |<ε, it is determined that trace segments i and j belong to the same trace. This is the Doppler consistency within the trace segment, where ε is the Doppler feature judgment threshold; ε can be set to a positive number close to 0. In this embodiment, based on the Doppler feature value sequence obtained in step 4-1-1, it can be concluded that trace segments S1 and S4 belong to the same trace P6, and trace segments S2 and S3 belong to the same trace P5.

[0142] Step 4-2: determine whether different traces belong to the same target based on the Doppler eigenvalues ​​between the traces. This is the Doppler similarity between the traces.

[0143] In step 4-3, the number of trace groups belonging to the same target is counted. The resulting number of trace groups is the target count result. In this embodiment, steps 4-1 and 4-2 are repeated until the Doppler eigenvalue classification results for all traces are obtained. Traces P1...P6 contain two different Doppler eigenvalues, belonging to two different targets, T1 and T2, respectively. Therefore, the number of targets in the non-line-of-sight area is 2.

[0144] Step 5: Use the trace order matching algorithm to determine the order of each multipath echo. Specifically:

[0145] Step 5-1: For each echo trace that has been determined to belong to a target, record the energy sum of each trace corresponding to each target in the time interval. Assuming that there are s traces corresponding to target b, calculate the energy sum of these s traces to obtain the energy sequence E b1 ,…,E bs , sort the energy sequence in descending order; according to the different echo reflection orders, classify the echo signals according to the reflection order and obtain the order matrix Ψ:

[0146]

[0147] Where i represents the number of reflections of the signal during transmission, and j represents the number of reflections of the signal during reflection. Since the diffraction energy of high-order scattering and high-frequency signals is extremely weak, the highest reflection order of the signal is only considered to be 3rd order;

[0148] Traverse all i,j values ​​and construct a multipath set consisting of different reflection times;

[0149] Step 5-2, select the trace with the largest energy as l 11 , select the energy and the second largest trace as l 12 , solve the target position through the ellipse cross positioning algorithm in step 6, and obtain the virtual target point T * , according to the virtual target point position and the radar position, combined with the multipath propagation model, the propagation distance of each order multipath corresponding to the virtual target point is obtained, and the difference d between the propagation distance corresponding to the virtual target point and the propagation distance obtained by the multipath propagation model is calculated;

[0150] Step 5-3: traverse all multipath sets, repeat step 5-2, and select the set with the smallest difference d as the trace order matching result. The target attribution determination and trace order information in this embodiment are shown in Table 1;

[0151] Table 1 Target attribution determination results and trace order information

[0152] trace Doppler eigenvalue (Hz) Target Trace order <![CDATA[P1]]> 1.25 <![CDATA[T1]]> <![CDATA[l 11 ]]> <![CDATA[P2]]> 1.50 <![CDATA[T2]]> <![CDATA[l 11 ]]> <![CDATA[P3]]> 1.50 <![CDATA[T2]]> <![CDATA[l 12 ]]> <![CDATA[P4]]> 1.50 <![CDATA[T2]]> <![CDATA[l 22 ]]> <![CDATA[P5]]> 1.25 <![CDATA[T1]]> <![CDATA[l 12 ]]> <![CDATA[P6]]> 1.50 <![CDATA[T2]]> <![CDATA[l 13 ]]>

[0153] In step 5-4, the order information of each trace corresponding to each target is recorded as the basis for subsequent target positioning and trajectory reconstruction.

[0154] Step 6: Using the ellipse cross positioning algorithm, combined with the multipath scattering model and the order information of each trace, the real position of the human target is determined and the motion trajectory is reconstructed, as shown in Figure 7 As shown in the figure, the experimental scene parameters include the wall size and radar placement position, which are d1 = 5.3m, d2 = 5.7m, d3 = 3.3m, and d4 = 7.0m. The lower left corner is the origin of the coordinate system, and the radar placement coordinates are (3.2m, 4.8m). Step 6 is as follows:

[0155] Step 6-1: Group the extracted traces according to their reflection order information, where the first-order reflection trace corresponds to the single-reflection echo, and the first- and second-order mixed reflection trace corresponds to the echo that reaches the target after the first reflection and then reaches the receiving end after the second reflection;

[0156] Step 6-2: For the first-order reflection trace, calculate the radial distance d1 from the target to the radar based on the echo propagation time t1 and the propagation velocity c1. The formula is:

[0157]

[0158] Step 6-3: For the first-order and second-order mixed reflection traces, an elliptical structure is formed between the target and the two reflecting walls. The total one-way propagation distance d of the first-order trace and the first-order and second-order mixed traces is * for:

[0159] d * =d r11 +d r12

[0160] Among them, d r11 with d r12 They represent the one-way distance of the first-order reflection trace and the one-way distance of the first- and second-order mixed reflection trace, d * is the major axis of the ellipse;

[0161] Step 6-4: Use d1 obtained in step 6-2 as the radius of the circle and the virtual radar coordinate point as the center of the circle to construct circle C1. * As the major axis of the ellipse, the first virtual radar coordinate point and the second virtual radar coordinate point are used as the two foci of the ellipse to construct the ellipse E1;

[0162] Step 6-5, solve the intersection of the first-order circle equation and the second-order ellipse equation as the estimated position of the target in the current time unit, and further connect the estimated positions of all time units to form a continuous motion trajectory of the target, such as Figure 8 As shown, the real position of target one is reciprocating within (4.6m, 7.0m)→(4.6m, 10.0m), and the real position of target two is stepping in place at (5.7m, 9.4m). The depth map is the positioning result of the present invention.

[0163] The above comparison demonstrates that the proposed method for tracking unknown multi-target motion trajectories in non-line-of-sight environments based on frequency-modulated continuous wave radar can accurately extract target echo traces in complex multipath environments, completing trace attribution, target counting, and trajectory reconstruction, validating the effectiveness and reliability of the proposed method. By introducing an improved trace extraction algorithm based on grid optimization and conditional constraints, combined with a spindle-shaped front trace energy suppression mechanism and a trace attribution and target counting method based on Doppler consistency and similarity, the proposed method effectively improves the accuracy of unknown multi-target trace extraction and attribution judgment. Target positioning and trajectory reconstruction are achieved through an elliptical intersection positioning method based on trace order matching, resulting in overall performance superior to existing technologies.

[0164] In one embodiment, a system for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario is provided. The system includes sequentially executing:

[0165] The first module is used to use a frequency modulated continuous wave radar to collect target echo signals in the non-line-of-sight area and pre-process the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing human target motion characteristics;

[0166] The second module is used to extract the target motion trajectory using an improved adaptive trajectory extraction algorithm based on gridding and conditional constraints. The gridding is to divide the time axis and distance axis into unit grids based on human body size and human motion characteristics. The conditional constraints are used to deal with trajectory jumps and trajectory breaks, including acceleration condition constraints and trajectory break tolerance constraints.

[0167] The third module is used to reduce the cross-interference between different traces through the spindle-shaped front trace energy suppression algorithm, and repeatedly execute the second and third modules until all traces are extracted;

[0168] The fourth module is used to complete the track segment ownership determination and target ownership determination based on the Doppler consistency within the track segment and the Doppler similarity between tracks, and simultaneously complete the target counting;

[0169] The fifth module is used to determine the order of each multipath echo through a trace order matching algorithm;

[0170] The sixth module is used to realize target positioning and reconstruct the target motion trajectory through the elliptical cross positioning algorithm combined with the multipath propagation model and trace order information.

[0171] Regarding the specific limitations of the system for tracking the motion trajectories of unknown multiple targets in non-line-of-sight scenarios, please refer to the limitations of the method for tracking the motion trajectories of unknown multiple targets in non-line-of-sight scenarios above, which will not be repeated here. The various modules in the above-mentioned system for tracking the motion trajectories of unknown multiple targets in non-line-of-sight scenarios can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0172] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following is achieved:

[0173] Step 1: Use a frequency modulated continuous wave radar to collect target echo signals in a non-line-of-sight area, and preprocess the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing human target motion characteristics;

[0174] Step 2: Extract the target motion trace using an improved adaptive trace extraction algorithm based on gridding and conditional constraints. The gridding algorithm divides the time axis and distance axis into unit grids based on human body size and motion characteristics. The conditional constraints are used to deal with trace jumps and trace breaks, including acceleration condition constraints and trace break tolerance constraints.

[0175] Step 3: Using the spindle-shaped front trace energy suppression algorithm to reduce cross-interference between different traces, repeat steps 2 and 3 until all traces are extracted;

[0176] Step 4: Based on the Doppler consistency within the trace segment and the Doppler similarity between traces, the trace segment ownership determination and target ownership determination are completed, and the target counting is completed at the same time;

[0177] Step 5: Determine the order of each multipath echo using a trace order matching algorithm;

[0178] Step 6: Target positioning is achieved and the target motion trajectory is reconstructed through the ellipse intersection positioning algorithm combined with the multipath propagation model and trace order information.

[0179] For the specific limitations of each step, please refer to the limitations of the method for tracking unknown multi-target motion trajectories in non-line-of-sight scenarios above, which will not be repeated here.

[0180] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the computer program implements:

[0181] Step 1: Use a frequency modulated continuous wave radar to collect target echo signals in a non-line-of-sight area, and preprocess the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing human target motion characteristics;

[0182] Step 2: Extract the target motion trace using an improved adaptive trace extraction algorithm based on gridding and conditional constraints. The gridding algorithm divides the time axis and distance axis into unit grids based on human body size and motion characteristics. The conditional constraints are used to deal with trace jumps and trace breaks, including acceleration condition constraints and trace break tolerance constraints.

[0183] Step 3: Using the spindle-shaped front trace energy suppression algorithm to reduce cross-interference between different traces, repeat steps 2 and 3 until all traces are extracted;

[0184] Step 4: Based on the Doppler consistency within the trace segment and the Doppler similarity between traces, the trace segment ownership determination and target ownership determination are completed, and the target counting is completed at the same time;

[0185] Step 5: Determine the order of each multipath echo using a trace order matching algorithm;

[0186] Step 6: Target positioning is achieved and the target motion trajectory is reconstructed through the ellipse intersection positioning algorithm combined with the multipath propagation model and trace order information.

[0187] For the specific limitations of each step, please refer to the limitations of the method for tracking unknown multi-target motion trajectories in non-line-of-sight scenarios above, which will not be repeated here.

[0188] Starting from the echo propagation characteristics in complex non-line-of-sight environments, the present invention makes full use of information such as distance, Doppler characteristics and echo reflection order, and proposes a lightweight, robust and high-precision unknown multi-target tracking method, which can effectively reduce the dependence on artificial experience parameters, while improving the universality and stability of the system in actual application scenarios.

[0189] The above describes the basic principles, main features, and advantages of the present invention. Those skilled in the art will appreciate that the present invention is not limited to the foregoing embodiments, and that the embodiments described in this specification are intended only to illustrate the basic concepts of the present invention. Any modifications, equivalent substitutions, and improvements to the present invention that do not depart from the spirit and scope of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for tracking unknown multi-target motion trajectories in a non-line-of-sight scenario, characterized in that: The method comprises the following steps: Step 1: Use a frequency modulated continuous wave radar to collect target echo signals in a non-line-of-sight area, and preprocess the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing human target motion characteristics; Step 2: Extract the target motion trace using an improved adaptive trace extraction algorithm based on gridding and conditional constraints. The gridding algorithm divides the time axis and distance axis into unit grids based on human body size and motion characteristics. The conditional constraints are used to deal with trace jumps and trace breaks, including acceleration condition constraints and trace break tolerance constraints. Step 3: Using the spindle-shaped front trace energy suppression algorithm to reduce cross-interference between different traces, repeat steps 2 and 3 until all traces are extracted; Step 4: Based on the Doppler consistency within the trace segment and the Doppler similarity between traces, the trace segment ownership determination and target ownership determination are completed, and the target counting is completed at the same time; Step 5: Determine the order of each multipath echo using a trace order matching algorithm; Step 6: Target positioning is achieved and the target motion trajectory is reconstructed through the ellipse intersection positioning algorithm combined with the multipath propagation model and trace order information.

2. The method for tracking unknown multi-target motion trajectories in a non-line-of-sight scenario according to claim 1, characterized in that: Step 1 specifically includes: Step 1-1: Use a frequency-modulated continuous wave radar to collect target echo signals in the non-line-of-sight area and rearrange the echo signals into an N×M matrix s(m,n). The vertical dimension of the rearranged data matrix is ​​the fast time dimension, and the horizontal dimension is the slow time dimension. The fast time dimension is divided according to the pulse sampling points, and the slow time dimension is divided according to the inter-pulse transmission interval. n represents the fast time sampling point index, and m represents the slow time transmission pulse index. Step 1-2, preprocessing the echo signal matrix, including DC component suppression and moving target detection; Steps 1-3: pre-process the echo matrix Perform fast Fourier transform to obtain the time distance graph matrix The calculation formula is: Where t is the time unit index, r is the distance unit index, j is the imaginary unit, R and T are the distance index range and time index range in the real number field, respectively.

3. The method for tracking unknown multi-target motion trajectories in a non-line-of-sight scenario according to claim 2, characterized in that: The DC component suppression in step 1-2 uses the mean filtering method, that is, the row mean is calculated along the slow time dimension of the echo signal matrix, and the row mean is subtracted from the original data to obtain the echo matrix after the DC component is suppressed. Specifically: Wherein, M is the slow time transmission pulse index.

4. The method for tracking unknown multi-target motion trajectories in a non-line-of-sight scenario according to claim 3, characterized in that: In step 1-2, the moving target detection is as follows: the echo matrix after DC component suppression is By offsetting and subtracting along the slow time axis, the stationary target echoes with the same amplitude and phase are eliminated, and the moving targets with continuously changing phases are retained to obtain the echo matrix after moving target detection. The calculation formula is as follows: Where n = [1, ..., N] represents the fast time sampling point, and m = [1, 2, ..., M] represents the slow time transmission pulse index.

5. The method for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario according to claim 2, characterized in that: Step 2 specifically includes: Step 2-1: On the time-distance diagram, grid the time and distance axes according to the spatial size of the human target. The time axis is divided into fixed time windows of length τ, where the value of τ is related to the inherent characteristics of human motion; the distance axis is divided into fixed distance intervals h, where h is related to human size. Gridding is performed along the fast time dimension of the time-distance diagram, using a grid size of τ × h and setting the grid sampling overlap based on the energy background information. Step 2-2: Based on the gridded time-distance graph, calculate the energy accumulation value of each grid in each time column to obtain the energy curve E corresponding to the current time column. n , and its calculation formula is: Among them, n' is the n'th grid index in the current time column, t represents the current indexed time column, i.e., the time unit index; N' is the total number of grids; Step 2-3: In each time window, the peak search method is used to find the energy curve E n Select the local maximum as the candidate grid r ij , specifically: R n ={r 1j ,r 2j ,…,r ij }=findpeaks(E ij ,E th ,h) Where i represents the i-th time column of the time distance graph, j represents the j-th grid of the i-th time column, and E ij is the current grid r ij Energy value, E th is the grid screening threshold, h is the human body size, and the selected grid r ij Store in grid matrix R n ; Steps 2-4: Apply conditional constraints on the candidate grid, specifically including: Step 2-4-1, introduce acceleration constraints, perform second-order differences on three consecutive grids along the time direction, and calculate the current grid r ij The trajectory acceleration a ij , whose expression is: Among them, τ' is the time interval between adjacent grids in the time dimension, are the extracted grids on the time columns (j-1)τ' and (j-2)τ' respectively; Step 2-4-2, apply energy penalty factor to the corresponding candidate trace point The updated grid energy is: Where λ is the acceleration penalty coefficient, which is used to control the smoothness of the trace, G is the updated grid matrix, and E[·] represents the energy value in the grid; r j is the jth time column of the current energy graph E, k represents the kth grid of the jth time column, E[r j ,k represents the energy value of the kth grid on the jth time column; Step 2-4-3, introduce the trace break tolerance constraint to deal with the trace break situation. When the break length n of the trace segment g Greater than the trace break tolerance threshold n th , delete the current trace; Step 2-5: On the updated grid matrix G, the trace finding problem is transformed into an optimization problem, and a set of grid points that maximizes the trace energy is extracted as the trace extraction result. Its expression is: Where N is the number of grids in the current j-th time column.

6. The method for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario according to claim 2, characterized in that: Step 3 specifically includes: Step 3-1: Starting from the center of the extracted trace grid point, a spindle-shaped energy suppression region is constructed. The suppression region is symmetrically extended in the time direction, and the energy in the grid decreases from the horizontal center line of the grid to both sides. Step 3-2: In the spindle-shaped inhibition region, the energy distribution of the original time-distance map Z(t,r) is corrected. The corrected energy Z * (t,d) is calculated according to the following formula: Among them, t and d represent the time column index and the grid index under the current time column in the time distance graph respectively. represents the extracted grid in time column j, and J0(·) is the zero-order Bessel function, specifically: Among them, x is used to represent variables and has no other meaning; Step 3-3, repeating steps 3-1 to 3-2, and performing spindle energy suppression processing on the extracted traces in sequence; Step 3-4: Repeat step 2 and step 3-1 to step 3-3 to extract all traces in the time-distance graph.

7. The method for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario according to claim 1, characterized in that: Step 4 specifically includes: Step 4-1: Extract the envelope curve of the short-time Fourier transform of the echo trace and perform a fast Fourier transform on the envelope curve to obtain the Doppler characteristic value of each echo trace. If there is any intersection of the traces, determine the trace segment attribution according to the following steps; Step 4-1-1, assume that there are q crossing trace segments near the intersection, with start and end times m and m respectively. s1 ,m s2 ,…,m sq With m e1 ,m e2 ,…,m eq , extract the envelope curve of the short-time Fourier transform of these q cross trace segments in their respective corresponding time intervals, and perform fast Fourier transform on the envelope curve to obtain the Doppler eigenvalue sequence of q echo trace segments, denoted as f d1 ,…,f dq ;m sq 、m eq are the start and end times of the qth cross trace segment, f dq is the Doppler eigenvalue of the qth cross-trace segment; Step 4-1-2, determine whether the cross trace segments belong to the same trace based on the Doppler eigenvalue. di -f dj When |<ε, the cross-trace segments i and j are considered to belong to the same trace, which is the Doppler consistency within the trace segment; where f di 、f dj are the Doppler characteristic values ​​of the i-th and j-th cross trace segments, respectively, and ε is the Doppler characteristic judgment threshold; Step 4-2: determine whether different traces belong to the same target based on the Doppler eigenvalues ​​between the traces. This is the Doppler similarity between the traces. In step 4-3, the trace sets belonging to the same target are counted, and the number of trace groups obtained is the target counting result.

8. The method for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario according to claim 1, characterized in that: Step 5 specifically includes: Step 5-1: For each trace that has been determined to belong to a target, record the energy sum of each trace corresponding to each target in the time interval. Assuming that there are s traces corresponding to target b, calculate the energy sum of these s traces to obtain the energy sequence E b1 ,…,E bs , sort the energy sequence in descending order; according to the different echo reflection orders, classify the echo signals according to the reflection order and obtain the order matrix Ψ: Among them, l ij Indicates that the number of reflections of the transmitted signal is i and the number of reflections of the reflected signal is j, i, j = 1, 2, 3; Traverse all i,j values ​​and construct a multipath set consisting of different reflection times; Step 5-2: In the energy sequence sorted in descending order, b1 As l 11 , with E b2 As l 12 , solve the target position through the ellipse cross positioning algorithm and obtain the virtual target point T * , according to the virtual target point position and the radar position, combined with the multipath propagation model, the propagation distance of each order multipath corresponding to the virtual target point is obtained, and the difference d between the propagation distance corresponding to the virtual target point and the propagation distance obtained by the multipath propagation model is calculated; Step 5-3, traverse all multipath sets, repeat step 5-2, and select the set with the smallest difference d as the trace order matching result; In step 5-4, the order information of each trace corresponding to each target is recorded as necessary information for subsequent target positioning and trajectory reconstruction.

9. The method for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario according to claim 8, characterized in that: Step 6 specifically includes: Step 6-1: Group the extracted traces according to their reflection order information, where the first-order reflection trace corresponds to the single-reflection echo, and the first- and second-order mixed reflection trace corresponds to the echo that first reaches the target after a single reflection and then reaches the radar sensor after a second reflection; Step 6-2: For the first-order reflection trace, calculate the radial distance d1 from the target to the radar based on the echo propagation time t1 and propagation velocity c1: Step 6-3: For the first-order and second-order mixed reflection traces, an elliptical structure is formed between the target and the two reflecting walls. The total one-way propagation distance d of the first-order trace and the first-order and second-order mixed traces is * for: d * =d r11 +d r12 Among them, d r11 with d r12 They represent the one-way distance of the first-order reflection trace and the one-way distance of the first- and second-order mixed reflection trace, d * is the major axis of the ellipse; Step 6-4, use d1 solved in step 6-2 as the radius of the circle, and the virtual radar coordinate point as the center of the circle to construct circle C1, and use d1 solved in step 6-3 as the radius of the circle. * As the major axis of the ellipse, the first virtual radar coordinate point and the second virtual radar coordinate point are used as the two foci of the ellipse to construct the ellipse E1; In step 6-5, the intersection of the first-order circle equation and the second-order ellipse equation obtained in step 6-4 is solved as the estimated position of the target in the current time unit. The estimated positions of the target in all time units are further connected in series to form a continuous motion trajectory of the target.

10. A system for tracking unknown multiple target motion trajectories in a non-line-of-sight scenario based on the method according to any one of claims 1 to 9, characterized in that: The system includes the following steps: The first module is used to use a frequency modulated continuous wave radar to collect target echo signals in the non-line-of-sight area and pre-process the echo signals, including data rearrangement, DC component suppression, moving target detection and Fourier transform, to obtain a time-distance map containing human target motion characteristics; The second module is used to extract the target motion trajectory using an improved adaptive trajectory extraction algorithm based on gridding and conditional constraints. The gridding is to divide the time axis and distance axis into unit grids based on human body size and human motion characteristics. The conditional constraints are used to deal with trajectory jumps and trajectory breaks, including acceleration condition constraints and trajectory break tolerance constraints. The third module is used to reduce the cross interference between different traces through the spindle-shaped front trace energy suppression algorithm, and repeatedly execute the second and third modules until all traces are extracted; The fourth module is used to complete the track segment ownership determination and target ownership determination based on the Doppler consistency within the track segment and the Doppler similarity between tracks, and simultaneously complete the target counting; The fifth module is used to determine the order of each multipath echo through a trace order matching algorithm; The sixth module is used to realize target positioning and reconstruct the target motion trajectory through the elliptical cross positioning algorithm combined with the multipath propagation model and trace order information.

Images