CN117192533A - Multipath ghost recognition method for vehicle-mounted millimeter wave radar based on reflection point inversion search - Google Patents

Abstract

The invention discloses a vehicle-mounted millimeter wave radar multipath identification method based on reflection point inversion search, which is applied to the technical field of vehicle-mounted millimeter wave multipath signal processing and aims at solving the problem that multipath ghosts in the prior art cannot be effectively identified and suppressed if the multipath ghosts cannot be effectively identified and suppressed; according to the method, the MIMO millimeter wave radar echo is preprocessed to generate Lei Dadian cloud; the method comprises the steps of performing dynamic and static point cloud separation on radar point clouds, clustering the separated dynamic point clouds, and acquiring center coordinates and multi-domain information of each cluster obtained by clustering; through target association, cluster centers in a scene are associated; calculating coordinates of reflection points under three assumptions; and preliminarily identifying multipath ghosts through a point cloud searching step, and finally further judging the multipath ghosts through a multi-domain feature.

Description

Multipath ghost recognition method for vehicle-mounted millimeter wave radar based on reflection point inversion search

Technical field

The invention belongs to the technical field of vehicle-mounted millimeter wave multipath signal processing, and particularly relates to a multipath ghost recognition technology.

Background technique

When using vehicle-mounted millimeter wave radar to sense the environment, the reflection and diffraction propagation of electromagnetic wave signals between different objects often cause complex multipath signals. When the multipath signal and the real target are in the same resolution unit in the range Doppler spectrum, the echo energy of the real signal will be reduced, resulting in missed detection; when the multipath signal and the real target are in the range Doppler spectrum, When located in different resolution units, multipath signals will form ghosts and cause false alarms. In autonomous driving scenarios, false alarms (multipath ghosts) caused by multipath signals are more common, and the interference they cause is more serious.

In recent years, domestic and foreign research institutions have carried out relevant technical research on the problem of multipath ghost interference in vehicle-mounted millimeter wave radar. In 2020, C. Liu et al. proposed a multipath ghost identification method based on small distance differences to solve the problem of multipath ghost interference (C. Liu, S. Liu, C. Zhang, Y. Huang and H .Wang, "Multipath propagation analysis and ghost target removal for FMCW automotive radars," IETInternational Radar Conference (IET IRC 2020), Online Conference, 2020, pp.330-334, doi:10.1049/icp.2021.0554.), however, this method The application scenario is very limited. When the target is close to the radar, the difference between the multipath ghost and the real target relative to the radar distance is large, and the above method cannot effectively identify the multipath ghost; and for adjacent lanes with similar distances , Two real targets with different angles are also prone to misrecognition. In 2023, Yunda Li and others proposed a multipath ghost identification method based on DOA/DOD estimation (Y.Li and X.Shang, "Multipath Ghost Target Identification for Automotive MIMO Radar," 2022IEEE96th Vehicular Technology Conference (VTC2022- Fall), London, United Kingdom, 2022, pp.1-5, doi:10.1109/VTC2022-Fall57202.2022.10012904.), by distinguishing the difference between the transmitting direction and the receiving direction of the antenna, and then identifying the differences between the transmitting direction and the incident direction. Multipath ghosting caused by electromagnetic propagation paths. However, the above method not only cannot solve the multipath ghosting when the transmitting and receiving directions are equal, but is also seriously affected by the antenna array.

In short, the above methods all have certain shortcomings and are not universally applicable. However, if multipath ghosts cannot be effectively identified and suppressed, the false alarms caused by multipath ghosts will seriously interfere with the detection of real targets in the scene, which is not conducive to the widespread application of automotive millimeter wave radar. Therefore, it is of great value to conduct in-depth research on multipath ghost interference identification methods for vehicle-mounted millimeter wave radars that are more universal and more effective.

Contents of the invention

In order to solve the above technical problems, the present invention provides a vehicle-mounted millimeter wave radar multipath ghost identification method based on reflection point inversion search.

The technical solution adopted by the present invention is: a vehicle-mounted millimeter wave radar multipath ghost identification method based on reflection point inversion search, including:

S1. Generate a radar point cloud based on the ADC data of the original echo of the current frame;

S2. Separate the dynamic and static point clouds of the radar point cloud, then cluster the separated dynamic point clouds, and finally extract the center coordinates of each cluster obtained by clustering and the multi-domain information characterizing the center coordinates;

S3. Correlate any two points from the cluster center coordinate set obtained in step S2 to form a matching set;

S4. Take a pair of related points from the matching set, perform inversion and solve based on the reflection points, and obtain the search area;

S5. Search whether the radar point cloud obtained in step S1 exists in the search area obtained in step S4. If it exists, perform step S6; otherwise, return to step S4;

S6. For the pair of related points processed in step S5, perform multipath ghost judgment based on multi-domain information;

S7. Traverse each pair of associated points in the matching set, repeat steps S4 to S6, and complete multipath ghost recognition of all point clouds under the current frame data.

Beneficial effects of the invention: The invention proposes a vehicle-mounted millimeter-wave radar multi-path ghost identification method based on reflection point inversion solution, which can effectively identify multi-path ghosts in vehicle-mounted millimeter-wave radar echoes. Through preprocessing of MIMO millimeter wave radar echo, radar point cloud is generated; through radar point cloud processing steps, dynamic and static point cloud separation, dynamic point cloud clustering, cluster center coordinates and multi-domain information acquisition are achieved; through target association, Correlate the cluster centers in the scene; calculate the reflection point coordinates under the three assumptions; initially identify multipath ghosts through the point cloud search step, and finally further determine the multipath ghosts through multi-domain characteristics. Actual measurement results show that the method of the present invention can effectively identify multipath ghosts in vehicle millimeter wave radar echoes.

Description of the drawings

Figure 1 is a schematic diagram of multipath signal propagation of a typical vehicle-mounted millimeter wave radar;

Figure 2 is an example diagram of the reference coordinate system for dynamic and static separation of point clouds;

Figure 3 is a schematic diagram of the geometric characteristics of different types of multipath ghosts;

Among them, (a) is a schematic diagram of the geometric characteristics of G _S , T, and P, (b) is a schematic diagram of the geometric characteristics of G _S , G _M2 , and T', (c) is a schematic diagram of the geometric characteristics of O, P, T, and G _M1 ;

Figure 4 is the experimental scene diagram;

Figure 5 shows the experimental data processing results;

Among them, (a) is the point cloud accumulated in multiple frames in the experimental scene, (b) is the moving target point cloud and multipath ghost, (c) is the multipath ghost recognition result, (d) is the multipath ghost elimination result.

Detailed ways

In order to facilitate those skilled in the art to understand the technical content of the present invention, the content of the present invention is further explained below with reference to the accompanying drawings.

Figure 1 is a schematic diagram of multipath signal propagation of a typical vehicle-mounted millimeter wave radar; the radar position is point O, the real target position is point T, and there is a reflection point in the scene as point P; generally speaking, the electromagnetic wave reflection is assumed to be specular reflection; therefore, the electromagnetic wave The transmission paths mainly fall into the following categories:

Direct-looking path: First, there is a two-way direct-looking path for detecting the target, O→T→O; secondly, there is also a two-way direct-looking path O→P→O caused by the backscattering of the reflection point P;

First-order multipath: The first-order path refers to the electromagnetic propagation path that passes through the reflection point P once. In the above scenario, there are two first-order multipaths, one is O→T→P→O, and the other is O→P→T →O;

Second-order multipath: Second-order multipath refers to the electromagnetic propagation path that passes through the reflection point P twice, specifically: O→P→T→P→O;

High-order multipath: High-order multipath refers to the electromagnetic propagation path that passes through the reflection point P multiple times. Generally, since electromagnetic reflection has a strong attenuation of signal energy, the high-order path leading to multiple reflections can be ignored because the signal energy is weak.

The relationship between electromagnetic propagation paths and detection targets is shown in Table 1:

Table 1 Relationship between electromagnetic propagation paths and detection targets

path name propagation path goals formed real target/ghost Path-0 O→T→O T real goal Path-1 O→P→O P real goal Path-2 O→P→T→O G _M1 ghost Path-3 O→T→P→O G _M2 ghost Path-4 O→P→T→P→O G _S ghost

The experimental scene is shown in Figure 4. The vehicle-mounted millimeter wave radar continuously emits electromagnetic wave signals into the scene, and the receiving antenna receives the echo signal. After mixing, filtering, and digital sampling, the original analog-to-digital converter (Analog-to-DigitalConverter, ADC) data. As shown in Figure 1, in a typical automotive millimeter-wave radar application scenario, the multipath ghosts generated by the reflection point P mainly include:

Type-1 secondary reflection multipath ghost G _M1 , the corresponding electromagnetic wave propagation path is O→P→T→O;

Type-2 secondary reflection multipath ghost G _M2 , the corresponding electromagnetic wave propagation path is O→T→P→O;

Type-2 triple reflection multipath ghost G _s , the corresponding electromagnetic wave propagation path is O→P→T→P→O;

Those skilled in the art should know that Type-1 in the present invention means that the electromagnetic wave is finally reflected from the real target back to the radar, and Type-2 means that the electromagnetic wave is finally reflected from the reflection point back to the radar. The arrows shown in Figure 1 represent the propagation of electromagnetic waves. direction.

The method of the present invention includes the following steps:

S1: signal preprocessing;

The signal preprocessing step realizes the generation process of the ADC data of the original echo to the radar point cloud. The main steps include fast time dimension fast Fourier transform, slow time dimension fast Fourier transform, constant false alarm detection, wave arrival angle estimation and coordinate conversion work. Assume that after processing one frame of radar ADC data, the generated radar point cloud set is P = {p ₀ , p ₁ , p ₂ ,..., p _NP }, where N _P represents the number of point clouds, p _i ={ x _i , y _i , vi _, RCS _i } represents the i-th radar point cloud and its multi-domain information.

Further, x _i , y _i , r _i , vi _, RCS _i respectively represent the abscissa coordinate of the radar point cloud in the automobile coordinate system, the ordinate coordinate of the radar point cloud in the automobile coordinate system, the distance of the radar point cloud relative to the radar, The radial velocity of the radar point cloud relative to the radar and the calculated radar cross-sectional area of the radar point cloud. The multi-frame point cloud accumulation result obtained from the preprocessing result is shown in Figure 5(a).

S2: Radar point cloud processing;

2.1 Separation of dynamic and static point clouds

The radar point cloud obtained in step S1 includes both point clouds of moving targets in the scene and point clouds of still objects in the scene. Through the point cloud separation step, the original radar point cloud is separated from static and moving parts to facilitate subsequent processing. Specifically, refer to the car coordinate system shown in Figure 2, with the car's forward direction as the y-axis and the horizontal direction perpendicular to the y-axis as the x-axis. Assume that the target is at the same height as the radar. Theoretically, the vehicle-mounted millimeter wave radar measures The radial velocity of a static object (stationary relative to the ground) for:

Among them, (x ^r , y ^r ) is the position of the radar in the car coordinate system, θ ^r is the angle between the normal direction of the radar and the car coordinate system, v _{C, x} , v _{C, y} and w respectively represent the movement of the car. Transverse and longitudinal velocity components and rotational speed. When the point cloud velocity measured by radar is the same as A large difference indicates that the point cloud is moving relative to the ground. Therefore, the basis for dynamic and static separation of point clouds is:

Among them, |·| means finding the absolute value, S and D respectively represent the static point cloud set and the dynamic point cloud set, and △v _sd is the speed threshold. In the present invention, the speed threshold △v _sd is taken as the empirical value of 0.8m/s. The result of separation of moving and static point clouds is shown in Figure 5(b). It can be seen from Figure 5 that in addition to the real point clouds representing moving targets, there are also many messy multipath ghosts.

2.2 Dynamic point cloud clustering

Road targets such as sports cars are extended for on-board millimeter wave radar. Therefore, the radar dynamic point cloud D needs to be clustered. Considering that the car is a rigid body moving target, the present invention adopts the DBSCAN clustering method based on speed improvement to perform clustering processing on the dynamic point cloud. Specifically, in the neighborhood search process of the core point of the traditional DBSCAN algorithm, further speed checks are performed on the points in the neighborhood of a core point, and points with a large speed difference from the speed of the core point are removed from the neighborhood of the core point. To exclude, in this embodiment, the Δv _DBSCAN value is set to 1m/s, that is, points with a speed difference greater than 1m/s from the core point are excluded from the neighborhood of the core point. The rest of the processing is the same as the traditional DBSCAN algorithm. In the present invention, the two important parameters of the DBSCAN algorithm are: neighborhood search radius Eps=2m, and density (the minimum number of points in the neighborhood) Minpts=1. In addition, the speed threshold used for core point speed check is taken as the empirical value Δv _DBSCAN =1m/s. Note that the result of a frame of radar data after clustering processing is CLU={clu ₀ ,clu ₁ ,clu ₂ ,...,clu _Ncluster }, where N _cluster represents the number of clusters after clustering, clu _i = {p _j ,...},p _j ∈D represents the i-th cluster.

2.3 Cluster center coordinates and multi-domain information acquisition

After the clustering process, the center coordinates of each cluster clu _i ={p _j ,...} are extracted, as well as the multi-domain information (speed, distance, RCS) characterizing the center point.

First, extract the center coordinates of the cluster. For general targets, we take the geometric center of the point cloud in the cluster as the center coordinates of the cluster (x _c , y _c ):

Among them, ( _xi ,y _i ) represents the coordinates of the point cloud in the cluster, and N represents the number of point clouds in the cluster. In particular, for large targets such as trucks, since the point cloud representing such targets spans a long space, the geometric center of the point cloud cannot well represent the parts of the target that cause multipath ghosting. Considering that parts with larger RCS have a greater probability of causing multipath ghosts, this invention extracts the center coordinates of clusters whose internal point cloud span exceeds 10m based on the RCS information of the point cloud. Specifically, first determine the maximum RCS calculated value of the point cloud in the i-th cluster. Then, search for the RCS calculated value and/> The point cloud set whose difference is within 10dBsm is recorded as clus′ _i ,/> Finally, the geometric center coordinates of the point cloud in clus′ _i are extracted as the center coordinates of cluster clu _i .

Furthermore, multi-domain features characterizing the cluster centers are extracted. This invention takes the Euclidean distance from the cluster center coordinates to the radar as the distance r _c from the cluster center to the radar, takes the average speed of the point cloud in the cluster as the speed v _c of the cluster center, and takes the maximum RCS value of the point cloud in the cluster as the cluster center. RCS, denoted as RCS _c .

Further, record the number Nup of point clouds in the cluster. Record the result of step 2.3 as Among them/>

S3: Target association, associate any two moving targets in the scene;

Cluster centers generated from S2 Take any two different points and associate them to form a matching set. Let the association operation be AS(·), then there is a matching set Γ:

Γ＝AS(S)

To reduce the amount of calculation, the above steps can be further optimized. It can be seen from the multipath characteristics that the relative radar distance difference between the real target and the multipath ghost is small. Therefore, target points within a certain distance difference can be selected for correlation to improve the real-time performance of the algorithm. Those skilled in the art should note that the certain distance difference here can take values of 5m, 10m and 15m respectively under the conditions of long-range radar, medium-range radar and short-range radar.

S4: Reflection point inversion solution;

Take a pair of associated points Γ ₀ = {c _A ,c _B }, Γ ₀ ∈ Γ in the matching set Γ. Among them, c _A = {x _a , y _a , r _a , v _a , RCS _a , Nup _a }, c _B = {x _b , y _b , r _b , v _b , RCS _b , Nup _b }. Since the distance from the multipath target to the radar is always longer than the distance from the real target to the radar, a further assumption can be made that r _a <r _b , that is, c _A is the real target and c _B is the multipath ghost. Assume the location of the radar is o(x _o , y _o ), and solve the reflection points under the three assumptions.

4.1 Solution of reflection points under the assumption of real ghost and Type-2 three-reflection multipath ghost G _S

As shown in Figure 3(a), the multipath ghost _GS and the real target are mirror symmetrical with respect to the reflection surface where the reflection point is located. Therefore, the intersection point of the mid-perpendicular line connecting the real target T and the multipath ghost _GS and the line connecting the multipath ghost _GS and the radar O is exactly the position of the reflection point. Assume that c _A is the real target T and c _B is the multipath ghost G _S. The specific process of solving the reflection point is:

First calculate the straight line

Further, calculate the line segment The mid-perpendicular line l _mid :A _mid x+B _mid y+C _mid ＝0:

Finally, the mid-perpendicular line l _mid is the same as the straight line The intersection point (x _bm ,y _bm ), that is, the theoretical coordinates of the reflection point P _TS under this assumption are:

As shown in Figure 3, the line segment Equivalent to the line segment where T and G _S are located in Figure 3(a);

Due to the influence of target measurement accuracy, signal noise and other factors, there are errors in the positioning of the target, which in turn affects the accuracy of the reflection point solution. Therefore, we set the reflection point search area to be an area with the reflection point P _TS as the center and a radius of △R, recorded as In this embodiment, the value of △R is 1m.

4.2 Solution of reflection points under the assumption of real ghost and Type-2 secondary reflection multipath ghost G _M2

As shown in Figure 3(b), the multipath ghost G _M2 is the midpoint of the line segment T'G _S. Therefore, the coordinates of T' can be obtained first, and then the second-order multipath ghost S can be determined according to the midpoint coordinate formula. Then use the method described in step 4.1 to determine the reflection point. Assume that c _A is the real target T and c _B is the multipath ghost G _M2 . The specific process of solving the reflection point is:

First calculate the line segment and line segments/> The included angle is:

Further, by rotating point c _A around point o by θ _∠AOB , the intersection point c _A′ can be obtained, which is T' in Figure 3(b). The calculation is as follows:

Furthermore, the multipath ghost G _S coordinates can be calculated according to the midpoint formula:

Here, use the method in step 4.1 to calculate the reflection point In the same way, the present invention sets the reflection point search area to reflect points/> is the center of the circle and the area with radius △R, denoted as/>

4.3 Solution of reflection points under the assumption of real ghost and Type-1 secondary reflection multipath ghost G _M1

As shown in Figure 3(c), the multipath ghost G _M1 and the real target have two characteristics. 1) The three points of the ghost G _M1 , the real target T and the radar coordinates are collinear; 2) The multipath ghost is caused by The reflection point of shadow G _M1 falls on an elliptical trajectory with point O and point T as the focus, and the sum of the distances from the two focal lengths is |OT|+|PT|. Assume that c _A is the real target T and c _B is the multipath ghost G _M2 . The specific process of solving the reflection point is:

First get the calculated line segment and line segments/> The angle θ _∠AOB . Due to the influence of radar angle measurement accuracy, noise and other factors, the multipath ghost G _M1 , the real target T and the reflection point P are not necessarily collinear after target detection. Therefore, we set a reasonable angle deviation threshold, and when the angle θ _∠AOB between the two line segments is lower than the threshold value Δε _θ , that is:

|θ _∠AOB |≤△ε _θ

It is considered that o, c _A and c _B are approximately three points collinear. In this embodiment, the value of Δε _θ is 1.5°.

Further, find the above elliptical trajectory. The parameters of the ellipse’s semi-major axis a _E , semi-minor axis b _E , and half-focal length c _E are respectively:

Furthermore, the angle between the ellipse and the positive direction of the x-axis is:

Finally, the elliptical trajectory can be expressed as:

Among them, (x _cc ,y _cc ) is a line segment the midpoint.

After obtaining the elliptical trajectory E, in the same way, we set the reflection point search area to be an elliptical annular area with the elliptical trajectory as the axis and the scaling factors as scale ₁ and scale ₂ , recorded as In this embodiment, the scaling factors scale ₁ and scale ₂ take values of 0.87 and 1.15 respectively.

When |θ _∠AOB |>△ε _θ , then this assumption is eliminated, let

S5: Based on the search domain solved in S4, search for reflection points in the dynamic and static point clouds to verify the hypothesis in S4.

After solving the reflection point area of S4, we can get three search areas Therefore, in the above-mentioned area, the moving and static point cloud set P obtained in step S1 is searched for whether there is a point cloud. If it exists, it can be preliminarily explained through the above geometric constraints that Γ ₀ = {c _A , c _B } is a result that falls in the search domain/> There are correlation point pairs between real targets and multipath ghosts caused by other strong RCS targets, and there are, c _A is the real target, and c _B is the multipath ghost.

S6: Multipath ghost judgment based on multi-domain information (distance, speed, RCS and point cloud number).

In order to improve the accuracy of multipath ghost recognition and avoid misidentification, multi-domain features are used to further determine multipath ghosts. It is known from step S4 that c _A ={x _a ,ya ,ra ,va ,RCS _{a ,} Nup _{a }} and c _B ={x _b ,y _b ,r _{b ,} v _b ,RCS _b ,Nup _b }. The decider refers to the following 3 factors:

Factor δ ₁ : Speed. There is a special relationship between the multipath ghost speed and the real target speed.

Among them, ε _v are speed thresholds respectively, which are taken as the empirical value 1m/s in this embodiment.

Factor δ ₂ : RCS. The reflection of electromagnetic waves will attenuate the echo energy, so the calculated RCS value of the real target will be greater than the calculated RCS value of the multipath ghost.

Factor δ ₃ : The number of point clouds. The number of point clouds representing the real target is not less than the number of point clouds of multipath ghosts corresponding to the real target. In this embodiment, the number of point clouds corresponding to the target is the number of point clouds in the cluster corresponding to the cluster center representing the target.

In summary, the multipath target determiner can be recorded as:

when The calculation result of is equal to 3, then it is determined that c _B is a multipath ghost; when/> The calculation result is less than 3, then it is determined that c _B is not a multipath ghost;

If a cluster center is identified as a multipath ghost, the point clouds in the corresponding cluster are all marked as multipath ghosts.

S7: Traverse each associated pair in the matching set obtained in step S3, repeat steps S4 to S6, and complete the multipath identification task of all point clouds under the current frame data.

The final recognition result is shown in Figure 5(c). The radar point cloud marked as a box is identified as a multipath ghost. As can be seen from Figure 5(c), after being processed by the proposed inventive technology, the multipath Ghosts can be effectively identified, and the results after eliminating multipath ghosts are shown in Figure 5(d). Actual measurement experiments have verified the feasibility and effectiveness of the present invention.

Those of ordinary skill in the art will appreciate that the embodiments described here are provided to help readers understand the principles of the present invention, and it should be understood that the scope of the present invention is not limited to such specific statements and embodiments. Various modifications and variations of the present invention may occur to those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of the claims of the present invention.

Claims (6)

1. The vehicle millimeter wave radar multipath identification method based on reflection point inversion search is characterized by comprising the following steps of:

s1, generating radar point cloud according to ADC data of an original echo of a current frame;

s2, performing dynamic and static point cloud separation on Lei Dadian clouds, then clustering the separated dynamic point clouds, and finally extracting the central coordinate of each cluster obtained by clustering and multi-domain information representing the central coordinate;

s3, any two points in the cluster center coordinate set obtained in the step S2 are taken for association to form a matching set;

s4, a pair of associated points are taken out from the matching set, and a search area is obtained based on inversion solution of the reflection points;

s5, searching whether the radar point cloud obtained in the step S1 exists in the search area obtained in the step S4, and if so, executing the step S6; otherwise, returning to the step S4;

s6, carrying out multipath ghost judgment on the pair of associated points processed in the step S5 based on multi-domain information, and if multipath ghosts exist in the pair of associated points, marking point clouds in the cluster corresponding to the multipath ghosts as multipath ghosts;

and S7, repeating the steps S4, S5 and S6 until the multi-path ghost identification task of all the point clouds under the current frame data is completed.

2. The method for identifying the vehicle millimeter wave radar multipath based on reflection point inversion search according to claim 1, wherein the central coordinates of the clusters are specifically as follows:

for the clusters corresponding to the targets with the point cloud span smaller than or equal to 10m, taking the geometric center of the point cloud to represent the cluster center of the clusters;

for the clusters corresponding to the targets with the point cloud spans exceeding 10m, the central coordinates of the clusters are specifically as follows:

a1, determining a calculated value of the maximum radar cross-sectional area of the point cloud in the cluster

A2, searching the calculated value of the radar sectional area in the clusterA point cloud with a difference within 10 dbm;

a3, extracting the geometric center coordinates of the point clouds in the point cloud set obtained in the step A2 to serve as the center coordinates of the cluster.

3. The method for identifying the vehicle-mounted millimeter wave radar multipath based on reflection point inversion search according to claim 2, wherein the multi-domain information comprises: the distance from the cluster center coordinates to the radar, the speed of the cluster center coordinates, the radar sectional area of the cluster center coordinates and the number of point clouds corresponding to the cluster center.

4. The method for identifying the vehicle millimeter wave radar multipath based on reflection point inversion search according to claim 3, wherein the reflection point inversion solution based on step S4 specifically comprises:

a pair of association points are taken out from the matching set, the distances between two points in the pair of association points and the radar are compared, the point with smaller distance is assumed to be a real target, and the other point is assumed to be multipath ghost;

respectively solving the coordinates of the reflection points corresponding to the following three assumed conditions, and obtaining respective search areas according to the coordinates of the reflection points:

assume case 1: the multipath ghost is specifically Typle-2 type tertiary reflection multipath ghost G _s ；

Assume case 2: the multipath ghost is specifically Typle-2 type secondary reflection multipath ghost G _M1 ；

Assume case 3: the multipath ghost is specifically Typle-1 type secondary reflection multipath ghost G _M2 。

5. The method for identifying the vehicle millimeter wave radar multipath based on reflection point inversion search according to claim 4, wherein step S6 is specifically:

s61, respectively calculating the following 3 factors:

velocity factor delta ₁ The calculation formula is:

RCS factor delta ₂ The calculation formula is:

point cloud number factor delta ₃ The calculation formula is:

s62, 3 causes calculated according to the step S61Sub-carrying out multipath ghost identification; specific: calculating a velocity factor delta ₁ RCS factor delta ₂ Point cloud count factor delta ₃ The sum of the three; if the calculated result is equal to 3, judging that multipath ghosts exist; otherwise, it does not exist.

6. The method for identifying multiple paths of vehicle millimeter wave radar based on reflection point inversion search according to claim 5, wherein when a certain cluster center is identified as multiple path ghosting, all point clouds in the cluster are marked as multiple path ghosting.