A target imaging method and system based on multipath echo suppression

Abstract

The present invention relates to the field of through-wall radar target imaging, and in particular to a target imaging method and system based on multipath echo suppression, the method comprising: obtaining the original echo of the through-wall radar; determining the pixel amplitude and the multipath echo peak in the target amplitude image according to the original echo and the multipath propagation coefficient using a delayed sum beamforming algorithm; performing echo elimination according to the pixel amplitude and the multipath echo peak to obtain multiple wall-eliminated echoes; fusing the multiple wall-eliminated echoes to determine a fused image; imaging the original echo using a BP algorithm to obtain an echo image; performing region division and sorting on the echo image in turn to obtain a pixel point set; screening overlapping targets according to the pixel point set, the amplitude threshold and the region limit ratio to obtain an overlapping target image; and determining the target image by superimposing the fused image and the overlapping target image. The present invention improves the accuracy of target imaging by eliminating part of the echo.

Description

Technical Field

[0001] The present invention relates to the field of through-wall radar target imaging, and in particular to a target imaging method and system based on multipath echo suppression. Background Art

[0002] In the process of through-the-wall radar detecting and imaging targets in a closed room, in addition to the clutter generated by the front wall, the multipath echoes reflected by the other walls will interfere with the accurate imaging of the target. Different from the general direct path echo, due to multiple reflections from obstacles such as walls, the target information in the multipath echo will be erroneous and eventually form a "false target". Multipath virtual images will cause false alarms in target imaging, seriously affecting subsequent target detection and recognition. Effective suppression of multipath echoes is an important prerequisite for through-the-wall imaging radar to accurately image targets behind walls.

[0003] Single image domain or echo domain multipath suppression algorithms, such as sub-aperture fusion and echo cancellation algorithms, often eliminate multipath virtual images by weighted mapping, multiplication fusion or echo cancellation of pixel amplitudes. However, when the target overlaps with the multipath virtual image, it is often impossible to completely suppress the virtual image and effectively retain the target at the same time, and the image signal-to-noise ratio and target imaging accuracy need to be improved. Summary of the invention

[0004] The object of the present invention is to provide a target imaging method and system based on multipath echo suppression, which can improve the accuracy of target imaging by eliminating part of the echoes.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A target imaging method based on multipath echo suppression, comprising:

[0007] Get the original echo of the through-wall radar;

[0008] Determine the pixel amplitude and multipath echo peak value in the target amplitude image according to the original echo and multipath propagation coefficient by using a delayed sum beamforming algorithm;

[0009] Perform echo elimination according to the pixel amplitude and the multipath echo peak value to obtain multiple wall-eliminated echoes;

[0010] Fusion of the plurality of wall echo-eliminated images to determine a fused image;

[0011] Imaging the original echo using a BP algorithm to obtain an echo image;

[0012] Sequentially dividing and sorting the echo images into regions to obtain a pixel point set;

[0013] Screening overlapping targets according to the pixel point set, amplitude threshold and area limit ratio to obtain overlapping target images;

[0014] The target image is determined by superimposing the fused image and the overlapping target image.

[0015] A target imaging system based on multipath echo suppression, comprising:

[0016] An acquisition module is used to obtain the original echo of the through-wall radar;

[0017] A delayed sum beamforming module, used to determine the pixel amplitude and multipath echo peak value in the target amplitude image according to the original echo and multipath propagation coefficient using a delayed sum beamforming algorithm;

[0018] An echo elimination module, used for performing echo elimination according to the pixel amplitude and the multipath echo peak value to obtain multiple wall-eliminated echoes;

[0019] A fusion module, used for fusing the plurality of wall echo cancellations to determine a fused image;

[0020] A BP algorithm imaging module, used for imaging the original echo using the BP algorithm to obtain an echo image;

[0021] A region division and sorting module, used to sequentially divide and sort the echo images into regions to obtain a pixel point set;

[0022] An overlapping target screening module is used to screen overlapping targets according to the pixel point set, the amplitude threshold and the area limit ratio to obtain an overlapping target image;

[0023] The superposition module is used to determine the target image by superimposing the fused image and the overlapping target image.

[0024] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0025] The present invention obtains the original echo of the through-wall radar; uses the delayed summation beamforming algorithm to determine the pixel amplitude and the multipath echo peak in the target amplitude image according to the original echo and the multipath propagation coefficient; performs echo elimination according to the pixel amplitude and the multipath echo peak to obtain multiple wall-eliminated echoes; fuses the multiple wall-eliminated echoes to determine the fused image; uses the BP algorithm to image the original echo to obtain an echo image; divides and sorts the echo image in turn to obtain a pixel set; screens overlapping targets according to the pixel set, the amplitude threshold and the area limit ratio to obtain an overlapping target image; and determines the target image by superimposing the fused image and the overlapping target image. The peak of the multipath echo is determined by the amplitude of each pixel and the multipath propagation coefficient, and the original echo is partially eliminated of each wall multipath echo through delay compensation and weighted operation, thereby improving the accuracy of target imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0027] Figure 1 This is a schematic diagram of the inner wall multipath;

[0028] Figure 2 This is a schematic diagram of the oscillation multipath in the front wall;

[0029] Figure 3 It is a schematic diagram of the interaction multipath between targets;

[0030] Figure 4 It is a schematic diagram of the multipath model;

[0031] Figure 5 It is a schematic diagram of the multipath equivalent calculation model;

[0032] Figure 6 It is a schematic diagram of the transmission of the two-layer medium model;

[0033] Figure 7 This is a schematic diagram of multipath target positioning;

[0034] Figure 8 This is a schematic diagram of multipath virtual image positioning;

[0035] Fig. 9 Schematic diagram of multipath virtual image and multipath target;

[0036] Fig.10 This is a schematic diagram of special scenario 1;

[0037] Fig.11 This is a schematic diagram of special scenario 2;

[0038] Fig.12 This is a schematic diagram of special scenario 3;

[0039] Fig.13 Schematic diagram of pixel division and rectangle division;

[0040] Fig.14 A flowchart of the target imaging method based on multipath echo suppression provided by the present invention in practical application;

[0041] Fig.15 This is a flow chart of the target imaging method based on multipath echo suppression provided by the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] The object of the present invention is to provide a target imaging method and system based on multipath echo suppression, which can improve the accuracy of target imaging by eliminating part of the echoes.

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] In order to better suppress multipath echoes, a multipath model for through-wall imaging is established for closed rooms. Based on the needs of target imaging, the multipath type with greater interference is determined; based on the assumption of mirror reflection, the multipath propagation delay and propagation coefficient are solved, and then the specific form of multipath echo is derived; through double-circle positioning and focusing analysis, the multipath virtual image is located. This multipath model is the basis for the theoretical derivation of the multipath suppression algorithm.

[0046] In a closed room, the target echo will be reflected multiple times by obstacles such as walls, ceilings or floors, and other targets, thus forming different types of multipath echoes. According to the types of obstacles that the multipath echoes pass through, they can be divided into four types, namely, inner wall multipath, front wall oscillation multipath, ceiling / floor multipath, and target-to-target interaction multipath. In order to facilitate theoretical analysis and algorithm proposal, and combined with the actual needs of target imaging, each type of multipath echo is discussed in turn.

[0047] Inner wall multipath is caused by the side walls or back walls of the room. It usually reflects on the wall surface or inside. Its propagation path is as follows: Figure 1 shown.

[0048] Based on the number of times the multipath echo is reflected by obstacles in the round-trip path, internal wall multipath can be further subdivided into the following categories:

[0049] 1. First-order multipath: First-order multipath refers to multipath echoes that are only reflected once by obstacles in the round-trip path of the echo. It is generally composed of a direct path and a first-order reflection path. Due to the small number of reflections and short propagation distance, the amplitude of the first-order multipath is similar to that of the direct path echo, constituting the main multipath echo.

[0050] 2. Second-order multipath: Second-order multipath refers to multipath echoes that are reflected twice by obstacles in the round-trip path of the echo. It can be divided into two cases: one is the combination of the direct path and the second-order reflection path, and the other is the combination of the first-order reflection path and the first-order reflection path. Generally speaking, only the second-order multipath along the same first-order reflection path is considered, and the other second-order multipath echoes have small amplitudes and are difficult to focus on imaging.

[0051] It is particularly noted that due to the long propagation distance, the virtual image of the second-order multipath is usually located outside the room range, which can be eliminated by constraining the imaging area or truncating the echo time dimension. At the same time, the signal attenuation caused by each wall reflection is considered, so the second-order multipath is ignored.

[0052] 3. High-order multipath: High-order multipath refers to multipath echoes that are reflected three or more times by obstacles in the round-trip path of the echo. Too many reflections make the amplitude of the multipath echo particularly weak, and the impact on the target imaging is very small. At the same time, the propagation path is complex, so high-order multipath is not included in the multipath model of the present invention.

[0053] The front wall oscillation multipath is a multipath that has multiple reflections in the front wall. Its propagation path is as follows: Figure 2 Generally speaking, the virtual image of the oscillation multipath in the front wall is located behind the target position, and since the multiple reflections in the wall greatly attenuate the signal, its amplitude is very weak, so this type of multipath can be ignored.

[0054] Ceiling / floor multipath is multipath caused by the ceiling or floor, similar to the inner wall multipath, and this type of multipath is also classified into first-order multipath, second-order multipath and high-order multipath. Considering that the through-wall imaging in the present invention is in the scope of two-dimensional imaging, and the ceiling / floor multipath needs to be discussed in the framework of three-dimensional imaging, this type of multipath is ignored.

[0055] Inter-target multipath is the multipath reflected between multiple targets. Its propagation path is as follows: Figure 3As shown. Generally speaking, this type of multipath will interfere with target imaging only in the imaging model of small scene and multiple targets. This is because the diffraction signal from irregular shaped targets and small targets has little effect; while large scatterers such as walls will scatter the diffraction waves of other targets in the same direction, forming interference signals with greater influence. In combination with the through-wall detection scenario in the present invention, this type of multipath is ignored.

[0056] In summary, the multipath model in the present invention only considers the first-order multipath of the inner wall, and the multipath virtual image formed by its focusing is the main interference to the target imaging.

[0057] Based on the clutter model, the imaging scene is expanded to obtain Figure 4 The multipath model shown in the figure. Where H is the length of the room distance, d w is the thickness of the front wall, L0 is the aperture of the antenna array and is the same width as the wall, h1 is the vertical distance between the antenna array and the wall, K is the number of array elements, P is the point target behind the wall, and O is the center position of the array. For the convenience of analysis, the following adjustments are made: without considering the internal reflection of the wall, the thickness of the left wall, the right wall and the back wall are ignored; the relative dielectric constant of each wall is set to ε r , ignoring the material differences of the walls; in order to better highlight the refraction details of each path in the front wall, the length representation in the figure does not follow a strict proportional relationship, and everything is based on the marked length. Figure 4 Path-D is the direct path between the point target P and the antenna array element, and Path-L, Path-R, and Path-B are the first-order reflection paths corresponding to the left wall, right wall, and back wall, respectively. Since the wavelength of the transmitted signal is greater than the roughness of the wall, it is assumed that the wall is smooth and specular reflection occurs here.

[0058] according to Figure 4 In the multipath model in , it is assumed that the through-wall imaging radar detects a single point target behind the wall with a co-located antenna array, and the antenna array is parallel to the wall. The echo signal received by the kth antenna array element can be expressed as

[0059]

[0060] Among them, s(t) is the transmitted signal, a tar (k, t, p) is the target echo, a m (k, t, p) is the multipath echo, C kpj , j∈{D,L,R,B} represents the one-way propagation coefficient between the target and the kth antenna element along Path-j, τ kpj ,j∈{D,L,R,B} represents the one-way propagation delay between the target and the kth antenna element along Path-j. kpDis the one-way propagation coefficient between the target and the kth antenna element along Path-D, s(t-τ kpD -τ kpj ) is the first-order multipath echo signal.

[0061] In order to determine the position of the multipath echo in the original echo, the multipath propagation delay must be calculated first. Figure 5 The figure shows the multipath equivalent calculation model. PL =(x pl ,y pl ), right wall mirror point X PR =(x pr ,y pr ), rear wall mirror point X PB =(x pb ,y pb );θ kpj , are the angles of incidence and refraction of each path respectively; is the reflection angle of the first-order multipath on each wall; X Lp , X Rp , X Bp They are the reflection points of the first-order multipath on each wall.

[0062] Next, take the first-order multipath of the left wall as an example to calculate its multipath propagation delay. According to the principle of mirror reflection, it is not difficult to find that the point target P and the mirror point P on the left wall L are mirror images of each other, then ||X P X Lp ||=||X PL X Lp ||. Then, the length of the first-order reflection path between the antenna array element and the real target is equal to the direct path length between it and the mirror target. Similarly, the corresponding propagation delays of the two are also equal. At this time, P L The coordinates can be expressed as (-L0-x p ,y p ), then the propagation delay of Path-L is

[0063]

[0064] Among them, y p is the ordinate of the point target P.

[0065] The accurate solution of the incident angle is usually achieved by numerical techniques such as Newton's method, which is a rather cumbersome process. Figure 5 The air-wall-air three-layer medium model in the paper is converted to Figure 6The wall-air two-layer medium model shown in the figure is used, and the approximate refraction compensation method is used to quickly calculate the propagation delay. It is particularly noted that the change of the medium model does not change the refraction angle and incident angle of the path.

[0066] Set point L kp0 Coordinate is X Lkp0 =(x lkp0 ,y lkp0 ), click L kp1 Coordinate is X Lkp1 =(x lkp1 ,y lkp1 ), click L kp2 Coordinate is X Lkp2 =(x lkp2 ,y lkp2 ). According to the approximate refraction compensation method, we have the equation set

[0067]

[0068] Among them, x k is the horizontal coordinate of the antenna array element. Solving equation (3), we can get the first-order multipath refraction point L on the left wall. kp0 The coordinates of At this time, the propagation delay of Path-L can be expressed as

[0069]

[0070] Similarly, for Path-R, Path-B, and Path-D, their propagation delays are

[0071]

[0072]

[0073]

[0074] When the signal propagates along each path, it will be refracted or reflected on the wall, which will cause the signal amplitude to change. The multipath propagation coefficient is used to describe this phenomenon. Here, we also take the first-order multipath of the left wall as an example to calculate its multipath propagation delay.

[0075] Horizontal polarization incident angle θ kpj , the reflection coefficient and refraction coefficient of j∈{D,L,R,B} are

[0076]

[0077]

[0078] Among them, Λ θkpj is the incident angle θ under horizontal polarizationkpj , the refractive index of j∈{D,L,R,B}, Γ θkpj is the incident angle θ under horizontal polarization kpj , the reflection coefficient of j∈{D,L,R,B}.

[0079] Depend on Figure 5 It can be seen that for Path-L, its propagation coefficient is

[0080]

[0081] Among them, Λ θkpL is the incident angle θ under horizontal polarization kpL The refractive index, is the refraction angle under horizontal polarization The refractive index, is the reflection angle under horizontal polarization The reflection coefficient, is the reflection angle of the first-order multipath signal on the left wall, θ kpL is the incident angle of the first-order multipath signal of the left wall on the front wall.

[0082] Similarly, for Path-R, Path-B, and Path-D, their propagation coefficients are

[0083]

[0084]

[0085]

[0086] To simplify the calculation, the real relative dielectric constant of the wall is used in equations (8) to (10), which does not affect the accuracy of the multipath model, but greatly improves the calculation efficiency. So far, the specific form of the multipath echo has been solved.

[0087] Taking the multipath virtual image of the left wall as an example, we examine the multipath target corresponding to a single antenna array element, such as Figure 7 As shown. Among them, the left wall mirror antenna array element X kL =(x kl ,0), multipath target X ML =(x ml ,y ml ), intersection point X AL =(x al ,y al ). For the convenience of analysis, the influence of the front wall on the signal propagation path is ignored and the position of the coordinate origin is adjusted.

[0088] According to the principle of mirror reflection, Figure 7 There is the following geometric relationship about multipath:

[0089] ||X P X k ||+||X P X AL ||+||X AL X k ||=||X PL X k ||+||X P X k || (14)

[0090] Among them, X P To point the target behind the wall, X k is the kth antenna array element.

[0091] From the perspective of propagation distance, the round-trip path length of the first-order multipath is equal to the sum of the lengths of the direct path and the first-order reflection path. In other words, the distance between the antenna array element and its multipath target can be expressed as

[0092]

[0093] Then, in the single-base working mode, the multipath target will follow a circular trajectory, and its trajectory equation is

[0094]

[0095] By introducing mirror antenna array elements, a bistatic radar is constructed. At this time, the following geometric relationship exists about multipath:

[0096] ||X PL X k ||+||X P X k ||=||X PL X k ||+||X PL X kL || (17)

[0097] Obviously, in the dual-base working mode, the multipath target will follow an elliptical trajectory, and its trajectory equation is

[0098]

[0099] Using double circular trajectories to jointly locate multipath targets, the intersection of the two can be obtained by combining the equations and

[0100]

[0101] Combination Figure 7 , discard the negative ordinate. It is not difficult to see that the multipath target always falls on the left wall and changes with the change of the antenna array element position.

[0102] like Figure 8 The figure shows the schematic diagram of multipath virtual image positioning. With antenna element T / R k-1 as the center, a multipath circular trajectory is drawn, which intersects the left wall at M'. L Point, intersecting with the multipath circular trajectory of T / R k at point.

[0103] A single multipath circular trajectory cannot provide the accurate position of the multipath virtual image. Only when the multipath circular trajectories of multiple antennas intersect can the multipath virtual image be located. Figure 8 As shown, the intersection of these multipath circular trajectories traversing the K array elements of the antenna array constitutes the positioning area of ​​the multipath virtual image. The precise solution of the multipath virtual image is very complicated, so we will not show it in detail here. We can try to solve it using the least squares method.

[0104] The previous discussion on the location of the multipath virtual image on the left wall is similar to the multipath virtual image on the right wall and the back wall. Fig. 9 As shown, there are multipath targets and multipath virtual images of each wall.

[0105] Generally speaking, no matter where the target is in the room, its multipath virtual image will usually fall near the wall, and the multipath target near the corresponding wall. This conclusion is particularly important for the theoretical analysis of multipath suppression.

[0106] After establishing the multipath model in a closed room, it can be clearly seen that the first-order multipath virtual image is the main interference to the target imaging, and the first-order multipath echo is the basis for the formation of multipath virtual images; at the same time, it is very difficult to directly locate the multipath virtual image in the image, but based on the directional dependence of the multipath virtual image, image fusion seems to have a good suppression effect.

[0107] Therefore, we start from both the echo domain and the image domain, firstly, we partially eliminate the multipath echoes of each wall from the original echo, and secondly, we fuse the differentiated images formed by various signals. In this way, we can effectively suppress the multipath virtual image.

[0108] like Fig.14 and Fig.15 As shown, the present invention provides a target imaging method based on multipath echo suppression, comprising:

[0109] Stage 1: Partial echo cancellation of multipath echoes

[0110] Through delayed summation beamforming, the target amplitude at each pixel position is obtained, which is multiplied by the multipath propagation coefficient to obtain the peak value of the corresponding multipath echo; in preparation for partial elimination of multipath echoes, the multipath echoes of each wall are partially eliminated from the original echo to obtain the echoes eliminated from each wall. Among them, the echoes eliminated from each wall are the echoes eliminated from the wall.

[0111] Step 101: Obtain the original echo of the through-wall radar.

[0112] Step 102: Determine the pixel amplitude and multipath echo peak value in the target amplitude image according to the original echo and multipath propagation coefficient using a delayed-sum beamforming algorithm.

[0113] Step 103: performing echo elimination according to the pixel amplitude and the multipath echo peak value to obtain multiple wall-eliminated echoes. Step 103 specifically includes: performing modulus weighting according to the pixel amplitude and the multipath echo peak value using a maximum function to obtain multiple wall-eliminated echoes.

[0114] Before eliminating multipath echoes, the first step is to use the existing original echoes to achieve an approximate representation of the multipath echoes. First, different from the general BP imaging algorithm, the original echoes are subjected to the following delayed summation beamforming processing. Then any pixel point X in the target amplitude image Q =(x q ,y q ) has a magnitude of

[0115]

[0116] Among them, τ kqD , C kqD They are pixel X Q Direct path propagation delay and propagation coefficient between antenna array elements. is the amplitude of any pixel in the target amplitude image, K is the number of array elements, k is the antenna array element, t is the time, a(k,t+2τ kqD ) is the pixel X Q The echo signal at .

[0117] Substituting formula (1) into formula (20), we can get

[0118]

[0119] Among them, L is the left wall, R is the right wall, B is the back wall, τ kpD is the one-way propagation delay between the target and the kth antenna element along Path-D, τ kpj is the one-way propagation delay between the target and the kth antenna element along Path-j, C kpD is the one-way propagation coefficient between the target and the kth antenna element along Path-D, C kpj is the one-way propagation coefficient between the target and the kth antenna element along Path-j, s(2τ kqD -τ kpD -τ kpj ) is the time delay of 2τ kqD -τ kpD -τkpj The echo signal, s(2τ kqD -2τ kpD ) is the time delay of 2τ kqD -2τ kpD echo signal.

[0120] For ease of analysis and without loss of generality, Fig. 9 As an example, the single target scenario shown in the figure is assumed to be The results of the formula are discussed as follows:

[0121] 1) Pixel X Q Coincident with point target P

[0122] At this time, τ kqD =τ kpD , C kqD =C kpD , τ kqD is pixel X Q The one-way propagation delay between the kth antenna element along Path-D, C kqD is pixel X Q The one-way propagation coefficient between the kth antenna element along Path-D is:

[0123]

[0124] 2) Pixel X Q Coincident with a certain multipath virtual image

[0125] Let's assume pixel X Q It coincides with the multipath virtual image of the left wall of the target, and there is 2τ kqD =τ kpD +τ kpL , C kpL is the one-way propagation coefficient between the target and the kth antenna element along Path-L, then

[0126]

[0127] 3) Pixel X Q It does not overlap with the point target P, nor with any multipath virtual image.

[0128] At this time, τ kqD ≠τ kpD , 2τ kqD ≠τ kpD +τ kpj ,j∈{L,R,B}, then

[0129]

[0130] From the above discussion, it is not difficult to see that in the processed image, the pixel points at the real target can maintain their original target amplitude, the pixel points at the multipath virtual image show a smaller target amplitude, and the target amplitudes of the remaining pixels are all 0. Overall, the pixel amplitude in the image can represent the target amplitude, and the "false target amplitude" at the multipath virtual image will be further discussed in subsequent processing.

[0131] Referring to formula (1), suppose a target is located at X Q At , its multipath echo can be expressed as

[0132] a mj (k,t,p)=C kqD C kqj s(t-τ kqD -τ kqj ),j∈{L,R,B} (25)

[0133] Among them, a mj (k, t, p) is the pixel X Q The multipath echo, C kqj is pixel X Q The one-way propagation coefficient between the kth antenna array element along Path-j.

[0134] The peak value of the echo is

[0135]

[0136] Among them, a′ mj (k,τ kqD +τ kqj ) is the echo peak value of the multipath echo.

[0137] Next, we need to locate and "subtract" this echo peak from the original echo. For any multipath target, the amplitude of its original echo signal is a(k,τ kqD +τ kqj ),j∈{L,R,B}, after subtracting the corresponding multipath echo peak, its signal amplitude is

[0138]

[0139] Since the amplitude of the signal is a complex number, direct subtraction will change the phase of the signal and affect coherent imaging. Therefore, the subtraction operation is replaced by modulus weighting. The signal amplitude at this time can be expressed as

[0140]

[0141] Here, special explanation is given for the "false target amplitude" at the multipath virtual image: tarIn (·), in addition to the real target, the pixel points at the multipath virtual image still have a certain target amplitude. Considering that this operation needs to traverse the pixel points, the signal amplitude at some positions will be negative. In order to avoid this phenomenon, the maximum value function is introduced to re-constrain the weighted value, which will not affect the elimination of multipath echoes. Then, the signal amplitude can be further written as

[0142]

[0143] Among them, a j (k,τ kqD +τ kqj ) is the signal amplitude after “subtracting” the peak value of the echo from the original echo, a(k,τ kqD +τ kqj ) is the original echo signal amplitude.

[0144] to i tar The pixel points in (·) are traversed, and the new echo signal is obtained as follows:

[0145]

[0146] Among them, the echo signal in formula (30) is the echo eliminated by the wall. j (k, t) is the echo eliminated from each wall, and a(k, t) is the original echo. Here, some letters are further explained. The letter a stands for amplitude, which is the abbreviation of amplitude. mj () indicates multipath echo. The echo is related to the return time. The formula is a' mj () represents the peak value of the multipath echo, so it has nothing to do with time and is C kqD ×C kqj ×itar(q); a() represents the amplitude of the original echo signal; a j () represents the signal amplitude of different paths after subtracting the corresponding multipath echo peak. itar(q) represents the amplitude of a certain point in the image. itar(·) represents the amplitude of all points in the entire image.

[0147] From the echo signal received by a single antenna array element, even under the above ideal conditions, the echo cancellation process has very high requirements on the accuracy of model parameters. As a result, it is difficult to effectively suppress multipath by relying solely on echo cancellation. Therefore, it is necessary to partially eliminate the multipath signal of the original echo to construct a differentiated image.

[0148] The so-called partial elimination means that the multipath echo of one wall is eliminated from the original echo each time, not all multipath echoes. The eliminated echo thus obtained retains the target echo and the multipath echoes of the other two walls. The original echo a(k, t) is partially eliminated of the multipath echo of the left wall to obtain the left wall eliminated echo a L (k, t); Similarly, the multipath echoes of other walls are partially eliminated in turn, and the right wall eliminated echo a is obtained respectively. R (k, t), back wall echo elimination a B (k,t).

[0149] Phase 2: Joint fusion of differentiated images

[0150] The BP algorithm is used to perform imaging processing on each wall to eliminate the echo, and each wall-eliminated image is obtained; and each wall-eliminated image is fused by joint multiplication to obtain a fused image.

[0151] Step 104: Fusing the plurality of wall-echo-eliminated images to determine a fused image. Step 104 specifically includes: imaging the plurality of wall-echo-eliminated images using a BP algorithm to obtain a plurality of wall-eliminated images; fusing the plurality of wall-eliminated images to obtain a fused image.

[0152] The directional dependence of multipath virtual images is an important property, that is, the position of multipath virtual images depends on the relative positions of the target, the wall and the antenna array elements. At the same time, no matter from which angle the target is detected, the position of the target remains unchanged. This statement is the theoretical basis for differential image fusion.

[0153] Eliminate echoes according to each wall a j (k, t), the BP algorithm is used to process the image, then each wall eliminates the pixel point X in the image Q The amplitude is

[0154]

[0155] in, Eliminate pixel X in the image for each wall Q The amplitude, a j (k,t+τ kqD +τ kqj ) is the signal amplitude in the image eliminated for each wall.

[0156] From the description of partial elimination of multipath echoes in the first stage, it is not difficult to see that in terms of signal amplitude alone, the echoes eliminated from each wall are at the same level. Therefore, the pixel amplitudes in each wall-eliminated image are also close to the original echo imaging. The so-called differentiated image only refers to the difference in individual multipath virtual images. Considering the multipath suppression effect and algorithm complexity, joint fusion is selected to process each wall-eliminated image.

[0157] Perform joint fusion on each wall elimination image, and fuse the pixel point X in the image Q The amplitude is

[0158]

[0159] in, is the pixel point X in the fused image Q The amplitude, Eliminate pixel X in the image for the left wall Q The amplitude, Eliminate pixel X in the image for the right wall Q The amplitude, Eliminate pixel X in the image for the back wall Q The amplitude.

[0160] It is not difficult to imagine that the target image in each image is in the same position, and the multipath virtual image can only appear in two of the images at most. In this way, the joint fusion of each image will effectively enhance the target imaging and better suppress the multipath virtual image. More importantly, the reliability of image fusion is very high, which can reduce the algorithm's dependence on model accuracy.

[0161] Phase 3: Overlapping Target Screening

[0162] The original echo imaging is divided into rectangular areas and sorted according to the amplitude of the pixel point sets; the first 10 pixel point sets are screened by amplitude and area limitation in turn to obtain overlapping target rectangular areas; and overlapping target images are generated using the pixels in the overlapping target rectangular areas.

[0163] Step 105: Using the BP algorithm to image the original echo to obtain an echo image.

[0164] Step 106: The echo image is divided into regions and sorted in sequence to obtain a set of pixel points. Step 106 specifically includes: determining rectangular region information according to the range resolution and azimuth resolution of radar imaging; dividing the region according to the echo image and the rectangular region information to obtain a set of rectangular images; and arranging the set of rectangular images in descending order according to the amplitude of the pixel points to obtain a set of pixel points.

[0165] Step 107: Screen overlapping targets according to the pixel point set, amplitude threshold and area limitation ratio to obtain overlapping target images.

[0166] Step 107 specifically includes:

[0167] The pixel point set is selected according to a set range to determine a screening set.

[0168] The filter set is filtered according to the amplitude threshold to obtain a target rectangle.

[0169] The target rectangles are screened according to the area limiting ratio to determine overlapping target rectangular areas.

[0170] An overlapping target image is generated according to the pixel points in the overlapping target rectangular area.

[0171] In the multipath model, there are several special target distribution scenarios:

[0172] 1) The target and its multipath virtual image overlap

[0173] like Fig.10 The figure shows the target distribution in this scenario. P1 is a point target and M1 is a multipath virtual image of the target. It can be seen that the two overlap and are both located near the wall.

[0174] 2) The target overlaps with the multipath virtual images of other targets

[0175] like Fig.11 The figure shows the target distribution in this scenario. P2 and P3 are point targets, and M2 and M3 are their corresponding multipath virtual images. It can be seen that P3 and M2 overlap each other and are both located near the wall. For the sake of generality, P3 and M3 do not overlap in the figure, but in most cases, P3, M2, and M3 overlap.

[0176] 3) The target overlaps with the virtual image of a "false target"

[0177] like Fig.12 The figure shows the target distribution in this scenario. P4 and P5 are point targets, M4 is the multipath virtual image corresponding to P4, and T1 is the multipath virtual image corresponding to the "false target" M4. In fact, P5 and M4 do not really overlap, but P5 is mistaken for the multipath virtual image of T1. This is different from the spatial overlap of the other two scenarios.

[0178] However, according to the analysis of the focus positioning of the multipath virtual image, the multipath virtual image generally appears near the wall. In any case, T1 must be close to the wall, and P5 at this time will also overlap with its own multipath virtual image.

[0179] In general, these three scenarios have one thing in common, that is, the target near the wall overlaps with its multipath virtual image. At this time, if the original echo is still processed by partial echo cancellation, the overlapping target will be mistaken for the multipath virtual image and eliminated, which is prone to target loss.

[0180] It is not difficult to find that the overlap of the target and its multipath virtual image will enhance the target imaging, and the pixel amplitude is relatively large in the original echo imaging. Therefore, it is advisable to find the overlapping target with the largest amplitude before echo elimination. In this way, no matter how it is processed later, the target will not be lost.

[0181] The original echo is imaged using the BP algorithm, and there are pixel points X Q The amplitude is

[0182]

[0183] The image i(·) is divided into Q = L1×L2 (azimuth × distance) pixels. From the perspective of grid division, the coordinates of any pixel are (m,n). Considering that a single pixel cannot form an effective target, a rectangular area of ​​a certain size is selected, and the pixels contained in it constitute a pixel set.

[0184] Assume that the image i(·) is re-divided into U×V (azimuth × distance) rectangles, each rectangle contains a×b (azimuth × distance) pixels, and the division is as follows: Fig.13 As shown, the amplitude of the pixel set with coordinates (u, v) is

[0185]

[0186] Among them, u is the azimuth coordinate of the rectangle, v is the distance coordinate of the rectangle, m is the azimuth coordinate of any pixel point, b is the number of distance pixels in the rectangular area, and n is the distance coordinate of any pixel point. It is the amplitude of each pixel in the traversal rectangular area.

[0187] It should be clear that the division of the rectangular area is directly related to the effect of target imaging. A rectangle that is too large may contain some useless pixels, interfering with target imaging; a rectangle that is too small will not be able to fully depict the target and will increase the amount of calculation. Taking into account the accuracy of target imaging and the amount of algorithm calculation, the range resolution δ1 and azimuth resolution δ2 of radar imaging are used to divide the rectangle together, and the value range of a and b is

[0188]

[0189]

[0190] Among them, χ1 and χ2 are the distance interval and azimuth interval of the pixel grid division respectively. For the sake of caution, only a reasonable reference range of rectangular division is given here, which can be appropriately selected according to actual needs.

[0191] All pixel sets in the image are sorted in descending order according to their magnitude, and the top 10 pixel sets are screened for the target rectangle. This process has the following two key points:

[0192] 1) Amplitude threshold of pixel set

[0193] According to the analysis of imaging principle, the pixel amplitude of overlapping targets can basically reach about twice that of ordinary targets. Let the amplitude of the i-th pixel set in amplitude sorting be i s (u i ,v i ),i=1,2,...,10, with i s (u1,v1) is used as the benchmark, and the amplitude of each pixel set is normalized.

[0194]

[0195] Setting the amplitude threshold like It can be preliminarily considered as the target rectangle, otherwise it is not. Considering the situation of multiple overlapping targets and the error of rectangular area division, it is allowed that there are multiple pixel points that meet the conditions and have similar amplitudes. Generally, it can be set

[0196] 2) Area limitation of pixel set

[0197] From the previous analysis, we can see that non-overlapping targets have been imaged in the fused image, and their repeated superposition is meaningless, but may reduce the image signal-to-noise ratio. In order to ensure that the selected pixel point set is an overlapping target, in addition to the amplitude requirement, the area limitation of the adjacent wall must also be met. In order to facilitate calculation and analysis, the area limitation ratios ι1 and ι2 are proposed, corresponding to the azimuth and distance restrictions respectively. At this time, the requirements for the pixel point set adjacent to the wall can be expressed as

[0198] {1≤u≤ι1U}∪{(U-ι1U+1)≤u≤U}∪{(V-ι2V+1)≤v≤V} (38)

[0199] Combining the rules of virtual image positioning and the imaging experience of actual scenes, the virtual image of the back wall is closer to the wall than the virtual image of the side wall. Generally, ι1=0.2 and ι2=0.1 can be set.

[0200] Only a set of pixels that meet both the amplitude and area requirements can be considered as the overlapping target rectangular area. Assume that the coordinates of the pixel set in this area are (u0, v0), then the pixel amplitude of the overlapping target image generated by this is

[0201]

[0202] Wherein, L1 is the number of pixels in the azimuth of image i(·), n0 is the range coordinate of the pixel in the image, m0 is the azimuth coordinate of the pixel in the image, u0 is the azimuth coordinate of the pixel set in the region, and v0 is the range coordinate of the pixel set in the region.

[0203] Obviously, for a scene with only non-overlapping targets, there is no set of pixels that meet the conditions, so the pixel amplitude of the overlapping target image is q=1,2,...,Q.

[0204] Stage 4: Image Overlay

[0205] The overlapping target image and the fused image are normalized respectively; the overlapping target image is superimposed on the fused image to obtain the final imaging result.

[0206] Step 108: Determine a target image by superimposing the fused image and the overlapping target image.

[0207] Step 108 specifically includes:

[0208] The pixel amplitudes of the fused image are normalized to obtain a normalized fused image; the pixel amplitudes of the overlapping target image are normalized to obtain a normalized overlapping target image; and the normalized overlapping target image is superimposed on the normalized fused image to obtain a target image.

[0209] Finally, the overlapping target images need to be superimposed on the fused image. In order to avoid the difference in amplitude between the two images affecting the target imaging effect, the pixel amplitudes of the two images are normalized separately.

[0210]

[0211]

[0212] Then, the two are superimposed to obtain the final image, whose pixel point X Q The amplitude is

[0213]

[0214] Aiming at the problem that the single echo domain or image domain multipath suppression algorithm is not thorough enough to suppress the virtual image and it is difficult to effectively retain the target, a multipath suppression algorithm based on partial echo elimination is proposed. In the framework of image fusion, the algorithm introduces the idea of ​​echo elimination, realizes the approximate representation of multipath echo based on the original echo and partially eliminates it, independently images the multipath elimination signal of each wall and performs joint fusion, and superimposes the possible missed targets through overlapping target screening to obtain the final imaging result.

[0215] The present invention also provides a target imaging system based on multipath echo suppression, comprising:

[0216] The acquisition module is used to obtain the original echo of the through-wall radar.

[0217] The delayed sum beamforming module is used to determine the pixel amplitude and the multipath echo peak value in the target amplitude image according to the original echo and the multipath propagation coefficient by using the delayed sum beamforming algorithm.

[0218] The echo elimination module is used to perform echo elimination according to the pixel amplitude and the multipath echo peak value to obtain multiple wall-eliminated echoes.

[0219] The fusion module is used to fuse the multiple wall echo eliminations to determine a fused image.

[0220] The BP algorithm imaging module is used to image the original echo using the BP algorithm to obtain an echo image.

[0221] The region division and sorting module is used to divide and sort the echo images in order to obtain a set of pixel points.

[0222] The overlapping target screening module is used to screen overlapping targets according to the pixel point set, amplitude threshold and area limitation ratio to obtain overlapping target images.

[0223] The superposition module is used to determine the target image by superimposing the fused image and the overlapping target image.

[0224] As an optional implementation, the echo elimination module specifically includes: an echo elimination unit, which is used to perform modulus weighting according to the pixel amplitude and the multipath echo peak value by using a maximum function to obtain multiple wall-eliminated echoes.

[0225] The present invention introduces the idea of ​​echo elimination within the framework of image fusion, and realizes effective suppression of multipath echoes through the multi-domain combination of image domain and echo domain. First, the peak value of the multipath echo is represented by the target amplitude and multipath propagation coefficient of each pixel point, and the multipath echo of each wall is partially eliminated from the original echo through time delay compensation and weighted operation. Secondly, the three types of signals obtained by processing are imaged independently, and a "clean" image can be obtained through joint fusion. In addition, considering the special case where the target overlaps with the virtual image, the target area with the largest amplitude near the wall is screened and superimposed on the fused image to obtain the final imaging result. Finally, the multipath suppression performance of the proposed algorithm is verified by simulation under various target distribution scenarios and comparison with the classic multipath suppression algorithm.

[0226] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0227] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A target imaging method based on multipath echo suppression, characterized in that: include: Get the original echo of the through-wall radar; Determine the pixel amplitude and multipath echo peak value in the target amplitude image according to the original echo and multipath propagation coefficient by using a delayed sum beamforming algorithm; Performing echo elimination according to the pixel point amplitude and the multipath echo peak value to obtain multiple wall-eliminated echoes, specifically comprising: using a maximum value function to perform modulus weighting according to the pixel point amplitude and the multipath echo peak value to obtain multiple wall-eliminated echoes; Fusion of the plurality of wall echo-eliminated images to determine a fused image; Imaging the original echo using a BP algorithm to obtain an echo image; Sequentially dividing and sorting the echo images into regions to obtain a pixel point set; Screening overlapping targets according to the pixel point set, amplitude threshold and area limit ratio to obtain overlapping target images; The target image is determined by superimposing the fused image and the overlapping target image.

2. The target imaging method based on multipath echo suppression according to claim 1, characterized in that: The step of eliminating echoes from the multiple walls and fusing them to determine a fused image specifically includes: Using the BP algorithm to image the plurality of wall echo eliminations to obtain a plurality of wall elimination images; A plurality of the wall-eliminated images are fused to obtain a fused image.

3. The target imaging method based on multipath echo suppression according to claim 1, characterized in that: The step of sequentially dividing and sorting the echo images into regions to obtain a set of pixel points specifically includes: Determine the rectangular area information according to the range resolution and azimuth resolution of the radar imaging; Performing region division according to the echo image and the rectangular region information to obtain a set of rectangular images; The rectangular image set is arranged in descending order according to the amplitudes of the pixel points to obtain a pixel point set.

4. The target imaging method based on multipath echo suppression according to claim 1, characterized in that: The screening of overlapping targets according to the pixel point set, the amplitude threshold and the area limit ratio to obtain the overlapping target image specifically includes: Selecting the pixel point set according to a set range to determine a screening set; Filter the filter set according to the amplitude threshold to obtain a target rectangle; The target rectangle is screened according to the area limiting ratio to determine the overlapping target rectangular area; An overlapping target image is generated according to the pixel points in the overlapping target rectangular area.

5. The target imaging method based on multipath echo suppression according to claim 1, characterized in that: Determining the target image by superimposing the fused image and the overlapping target image specifically includes: Normalizing the pixel amplitudes of the fused image to obtain a normalized fused image; Normalizing the pixel amplitudes of the overlapping target image to obtain a normalized overlapping target image; The normalized overlapping target image is superimposed on the normalized fused image to obtain a target image.

6. A target imaging system based on multipath echo suppression, characterized in that: include: An acquisition module is used to obtain the original echo of the through-wall radar; A delayed sum beamforming module, used to determine the pixel amplitude and multipath echo peak value in the target amplitude image according to the original echo and multipath propagation coefficient using a delayed sum beamforming algorithm; An echo elimination module is used to perform echo elimination according to the pixel amplitude and the multipath echo peak value to obtain multiple wall-eliminated echoes; the echo elimination module specifically includes: an echo elimination unit, which is used to perform modulus weighting according to the pixel amplitude and the multipath echo peak value using a maximum value function to obtain multiple wall-eliminated echoes; A fusion module, used for fusing the plurality of wall echo-eliminated signals to determine a fused image; A BP algorithm imaging module, used for imaging the original echo using the BP algorithm to obtain an echo image; A region division and sorting module, used for sequentially dividing and sorting the echo image into regions to obtain a pixel point set; An overlapping target screening module is used to screen overlapping targets according to the pixel point set, the amplitude threshold and the area limit ratio to obtain an overlapping target image; The superposition module is used to determine the target image by superimposing the fused image and the overlapping target image.

Images