Radar target track speed acquisition method and device, and storage medium

Abstract

The application discloses a radar target track speed acquisition method and device, and a storage medium. The method comprises the following steps: acquiring position coordinates of a plurality of detection points related to a target track; determining a heading angle of the target based on the position coordinates of the plurality of detection points; determining a radial speed of the target according to signal-to-noise ratios of the plurality of detection points; and obtaining a speed vector of the target based on the heading angle and the radial speed, so that the interference of inconsistent radial speeds of multiple parts of a non-rigid target on the acquisition of the speed vector of the target can be reduced, and the reliability of the speed vector of the target can be improved.

Description

Technical Field

[0001] This application relates to the field of radar testing technology, and in particular to a method and apparatus for obtaining radar target trajectory velocity, and a storage medium. Background Technology

[0002] In vehicle-mounted millimeter-wave radar systems, target trajectory initialization is a crucial step in achieving stable and reliable tracking. Especially in the fields of advanced driver assistance systems (ADAS) and autonomous driving, the accuracy of estimating the target's direction of motion and speed in the initial stage of the trajectory (i.e., the "trackhead" state) directly affects the speed of subsequent target tracking convergence, trajectory maintenance capability, and the overall system's safety and robustness.

[0003] Currently, the mainstream method for obtaining radar target trajectory velocity mainly relies on angle-radial velocity fitting technology. This technology acquires azimuth angle and radial velocity data pairs from multiple track-related detection points, and uses Random Sample Consensus (RANSAC) or least squares method to fit and calculate the target's velocity. It then calculates the target's heading angle and confirms the validity of the target's velocity when the fitting results meet quality indicators such as the number of inliers, the proportion of inliers, and the angle span.

[0004] Furthermore, when the aforementioned fitting method fails, existing technologies also assume the target's velocity V along the radar's lateral direction. y =0, and take the mean or maximum radial velocity of the detection points related to the trajectory as the target's longitudinal velocity V. X To obtain the target's trajectory speed (V) X ,0).

[0005] However, the aforementioned existing technology has the following problems:

[0006] 1) Sensitive to azimuth measurement errors: When the target is located at the edge of the radar's field of view (FOV), the azimuth measurement error increases significantly due to antenna beam distortion and decreased angular resolution, reaching several degrees or even more than ten degrees. Since methods such as RANSAC directly use the azimuth as an input variable for fitting, this error is directly transmitted to the velocity estimation result, causing the calculated target velocity to deviate significantly from the true value, especially in the large-angle region, where the rate of change of sin(θ) increases sharply, resulting in a significant deviation of the lateral velocity V. y The estimates are highly unstable.

[0007] 2) Sensitive to radial velocity dispersion: For non-rigid targets (such as motorcycles, bicycles, pedestrians, etc.), different parts (such as the vehicle body, wheels, rider's limbs, etc.) have different velocity vectors, resulting in highly discrete radial velocities at each detection point. Methods such as RANSAC implicitly assume that all detection points share the same velocity, i.e., the target is a rigid body moving as a whole. This assumption is severely distorted in non-rigid scenarios, leading to an excessively low proportion of interior points and fitting failure.

[0008] 3) Strict requirements on the number and spatial distribution of detection points: Methods such as RANSAC typically require at least 5 effective detection points with an angular span of no less than 5 degrees to initiate the fitting process. However, in the edge region of the radar field of view, targets can often only be detected at 3 to 4 points, making it difficult to meet the activation conditions.

[0009] 4) High computational complexity: In the worst case, the RANSAC algorithm requires up to 999 iterations. Each iteration involves sampling, model calculation and interior point determination, resulting in high computational complexity and potentially introducing significant latency on embedded real-time processing platforms.

[0010] 5) When fitting fails, the directional assumption is too rigid: the pure longitudinal motion assumption method completely ignores the possibility of lateral motion. When the target enters the radar field of view laterally at a large angle, this assumption will introduce a serious directional error, resulting in incorrect initial velocity orientation of the track. Incorrect initial velocity will cause the Kalman filter to converge slowly or even diverge, and in severe cases, it will cause track loss or miscorrelation.

[0011] Therefore, a new method for obtaining radar target trajectory velocity needs to be proposed to solve the above problems. Summary of the Invention

[0012] In view of the above problems, the purpose of this application is to provide a radar target trajectory velocity acquisition method, device and storage medium, which can reduce the interference of inconsistent radial velocities of multiple parts of non-rigid targets on the target velocity vector calculation and improve the reliability of the target velocity vector.

[0013] According to a first aspect of this application, a method for obtaining the velocity of a radar target track is provided, comprising: obtaining the position coordinates of a plurality of detection points related to the target track; determining the heading angle of the target based on the position coordinates of the plurality of detection points; determining the radial velocity of the target based on the signal-to-noise ratio of the plurality of detection points; and obtaining the velocity vector of the target based on the heading angle and the radial velocity.

[0014] Optionally, determining the heading angle of the target based on the position coordinates of multiple detection points includes calculating a covariance matrix based on the position coordinates of the multiple detection points; performing eigenvalue decomposition on the covariance matrix to obtain the eigenvector corresponding to the largest eigenvalue; and determining the heading angle based on the eigenvector.

[0015] Optionally, obtaining the target's velocity vector based on the heading angle and the radial velocity includes determining whether the direction of the feature vector is the same as the target's actual direction of motion based on the target's relative motion direction and track state; if the determination result is yes, using trigonometric functions to synthesize the velocity vector from the heading angle and the radial velocity; and if the determination result is no, using trigonometric functions to synthesize the velocity vector from the heading angle and the radial velocity after being flipped by 180°.

[0016] Optionally, determining the radial velocity of the target based on the signal-to-noise ratio of the plurality of detection points includes obtaining the signal-to-noise ratio of the plurality of detection points and comparing the signal-to-noise ratios of the plurality of detection points; and selecting the radial velocity of the detection point with the largest signal-to-noise ratio as the radial velocity of the target.

[0017] Optionally, determining the radial velocity of the target based on the signal-to-noise ratio of the plurality of detection points includes obtaining the signal-to-noise ratio of the plurality of detection points, determining the weight of the radial velocity of each detection point based on the signal-to-noise ratio of the plurality of detection points, and performing a weighted average of the radial velocities of the plurality of detection points to obtain the radial velocity of the target, wherein the weight of the radial velocity of each detection point is positively correlated with its signal-to-noise ratio.

[0018] Optionally, the method further includes filtering the velocity vector. The filtering is implemented using a Kalman filter. Before filtering, the state error covariance matrix of the Kalman filter needs to be initialized so that the initial error variance of the lateral velocity component is greater than the initial error variance of the longitudinal velocity component.

[0019] Optionally, determining the heading angle of the target based on the position coordinates of the plurality of detection points includes: linearly fitting the position coordinates of the plurality of detection points to determine the heading angle based on linear regression or Hall transform; or obtaining a weighted covariance matrix based on the position coordinates of the plurality of detection points and determining the heading angle based on the weighted covariance matrix using a principal component analysis algorithm; or establishing a point set based on the position coordinates of the plurality of detection points, detecting the convex hull of the point set, and extracting the principal axis from the convex hull to determine the heading angle.

[0020] According to a second aspect of this application, a method for acquiring radar target track velocity is provided, comprising: acquiring the azimuth and radial velocities of multiple detection points related to the target track when the target's trackhead state is updated; acquiring a first velocity vector of the target using a first track velocity acquisition method; updating the first velocity vector to the target's velocity vector when the first velocity vector is valid; and acquiring a second velocity vector of the target using a second track velocity acquisition method when the first velocity vector is invalid, and the track age reaches the maximum trackhead threshold and / or the target is at the edge of the radar field of view, and updating the second velocity vector to the target's velocity vector, wherein the second track velocity acquisition method employs the method described above.

[0021] Optionally, the method further includes obtaining the target's third velocity vector using a third trajectory velocity acquisition method when the first velocity vector is invalid, the track age reaches the maximum track head threshold, and the target is not at the edge of the radar field of view, and updating the third velocity vector to the target's velocity vector.

[0022] According to a third aspect of this application, a radar target track velocity initialization device is provided, comprising a memory for storing computer instructions; and a processor for implementing the radar target track velocity acquisition method as described above when executing the computer instructions.

[0023] According to a third aspect of this application, a storage medium is provided that stores a computer program, which, when executed by a processor, implements the radar target track velocity acquisition method as described above.

[0024] The radar target trajectory velocity acquisition method, apparatus, and storage medium provided in this application use the radial velocity of the detection point with the highest signal-to-noise ratio as the radial velocity of the target, or use the weighted average of the radial velocities of multiple detection points (the weight of the radial velocity of each detection point is positively correlated with its signal-to-noise ratio) as the radial velocity of the target. This makes the radial velocity used for radar target trajectory velocity initialization closer to the actual radial velocity of non-rigid targets, thereby avoiding the interference of inconsistent radial velocities of multiple parts of non-rigid targets on direction estimation and improving the reliability of the obtained velocity vector.

[0025] Furthermore, in an optional embodiment of this application, the target's heading angle is calculated using a robust principal component analysis algorithm. This ensures that the velocity vector obtained from the heading angle balances robustness and reliability, thereby improving the effectiveness of velocity initialization for non-rigid targets (e.g., even when the target is moving laterally or obliquely, an effective velocity vector can still be obtained). In addition, the principal component analysis algorithm not only requires fewer iterations but also has no requirements on the angle span. Theoretically, it only requires a minimum of two detection points to calculate the covariance matrix, thus enabling it to operate normally even in radar field-of-view edge scenarios with only 3-4 detection points.

[0026] Furthermore, in optional embodiments of this application, different trajectory speed acquisition methods are selected according to different scenarios to calculate the target velocity vector, so that the optimal velocity vector of the target can be obtained in different application scenarios, achieving full scenario coverage. Attached Figure Description

[0027] The above and other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0028] Figure 1 A schematic flowchart of a radar target trajectory velocity acquisition method according to an embodiment of this application is shown;

[0029] Figure 2 A flowchart illustrating another radar target trajectory velocity acquisition method according to an embodiment of this application is shown;

[0030] Figure 3 This diagram illustrates how the principal component analysis algorithm is used to obtain the target heading angle in an embodiment of this application. Detailed Implementation

[0031] Various embodiments of this application will now be described in more detail with reference to the accompanying drawings. In the various drawings, the same elements or modules are indicated by the same or similar reference numerals. For clarity, the various parts in the drawings are not drawn to scale.

[0032] It should be understood that, in the following description, "circuit" may include single or combined hardware circuits, programmable circuits, state machine circuits, and / or elements capable of storing instructions executed by the programmable circuit. When an element or circuit is said to be "connected" to another element or "connected" between two nodes, it may be directly coupled or connected to the other element, or there may be intermediate elements; the connection between elements may be physical, logical, or a combination thereof. Conversely, when an element is said to be "directly coupled to" or "directly connected" to another element, it means that there are no intermediate elements between them.

[0033] Furthermore, certain terms are used in this patent specification and claims to refer to specific components. Those skilled in the art will understand that hardware manufacturers may use different names to refer to the same component. This patent specification and claims do not distinguish components based on differences in name, but rather on differences in function.

[0034] Furthermore, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0035] Before introducing the technical solutions of the embodiments of this application, let's first introduce the professional names involved in the embodiments of this application:

[0036] 1) Principal Component Analysis (PCA): A statistical method that transforms potentially correlated variables into linearly uncorrelated variables through orthogonal transformation, with the new variables ordered in descending order of variance. In this application, it is used to analyze the principal axis direction of the spatial distribution of radar detection points.

[0037] 2) Track Header (TRACK_STATE_HEADER): The initial stage of the track (age less than the maturity standard). At this time, the track has only a few detection points, and the speed information has not yet been established or is extremely inaccurate.

[0038] 3) Track Initialization: The process of transitioning a track from a track head state to a mature track state, the core of which is to establish the target's position and velocity state estimation.

[0039] 4) Radial Velocity (V_r): The relative velocity component of the radar detection point along the radar line of sight, obtained by Doppler frequency shift measurement.

[0040] 5) Azimuth Angle (θ / azm): The horizontal angle of the radar detection point relative to the radar line of sight.

[0041] 6) Longitudinal / Lateral Velocity (V) x / V y ):V x V represents the velocity component along the radar line of sight (Range direction). yThe velocity component is perpendicular to the line of sight (Cross-Range direction).

[0042] 7) Covariance Matrix: A matrix that describes the correlation between the dimensions of a multidimensional random variable. Its diagonal elements are the variances of each dimension, and its off-diagonal elements are the covariances.

[0043] 8) Eigenvalue Decomposition: A mathematical operation that decomposes a square matrix into a set of eigenvectors and their corresponding eigenvalues. Eigenvectors indicate the direction of data variation, and eigenvalues ​​indicate the magnitude of variation in that direction.

[0044] 9) Field of View Edge (FOV Edge): The edge region of the radar's effective detection range, where the noise in angle measurement increases significantly.

[0045] 10) Non-rigid targets: Targets with multiple relatively moving parts (such as motorcycle wheels or human limbs), whose radar radial velocities at each part exhibit high dispersion.

[0046] 11) Association Detection Recorder (AssocDetRec): A historical association detection point data storage module for track maintenance, containing fields such as signal-to-noise ratio (SNR), azimuth, radial velocity, and position coordinates (x, y) of the detection points.

[0047] 12) Random Sample Consensus (RANSAC): A robust method for estimating mathematical model parameters through iterative sampling, which has good tolerance for outliers.

[0048] 13) Direction Ambiguity: The eigenvectors of PCA only determine the direction axis (straight line) and do not distinguish between positive and negative directions (180-degree ambiguity), which needs to be eliminated by additional criteria.

[0049] 14) Track Age: Track age is the cumulative number of radar scan frames that have been successfully associated with valid detection points since the track was first established. The track age increases by 1 for each valid detection point frame that is successfully associated with the track. When valid detection points are lost consecutively, the track age remains unchanged and only the number of frames lost to be associated is accumulated.

[0050] Figure 1A schematic flowchart of a radar target track velocity acquisition method according to an embodiment of this application is shown.

[0051] See Figure 1 The radar target trajectory speed acquisition method provided in this application includes steps S01-S04.

[0052] In step S01, the position coordinates of multiple detection points related to the target trajectory are obtained.

[0053] For example, the position coordinates of multiple detection points associated with the target track in at least one radar scan frame can be read from the Association Detection Recorder (AssocDetRec).

[0054] The location coordinates of each detection point include the horizontal coordinate x. i and the vertical coordinate y i , i=1,……,n. Where n is a positive integer, representing the number of detection points.

[0055] In step S02, the heading angle of the target is determined based on the position coordinates of multiple detection points.

[0056] For example, the heading angle of a target can be obtained using Principal Component Analysis (PCA). Specific methods include:

[0057] The covariance matrix is ​​calculated based on the position coordinates of multiple detection points, and eigenvalue decomposition is performed on the covariance matrix to obtain the largest eigenvalue and its corresponding eigenvector. Based on the eigenvector, the second heading angle of the target is calculated using the principal component analysis algorithm.

[0058] Specifically, the covariance matrix C is a 2×2 covariance matrix, which can be represented as: .

[0059] in,

[0060] var(X) represents the horizontal coordinates x of multiple detection points. i The variance;

[0061] var(Y) represents the vertical coordinates y of multiple detection points. i The variance;

[0062] cov(X, Y) represents the horizontal coordinates x of multiple detection points. i Vertical coordinate y i The degree of coupling of measurement errors is calculated using the following formula:

[0063] .

[0064] Where, xi The y-coordinate represents the horizontal coordinate of the detection point. i Indicates the longitudinal coordinates of the detection point. Represents the horizontal coordinate x of n detection points i The average value, Represents the vertical coordinates y of n detection points i The average value.

[0065] Furthermore, the decomposition of eigenvalues ​​includes:

[0066] t=C xx +C yy ;

[0067] d=C xx *C yy -C 2 xy ;

[0068] Δ=t 2 -4d;

[0069] Maximum eigenvalue ;

[0070] Minimum eigenvalue .

[0071] The methods for calculating the eigenvector corresponding to the largest eigenvalue include:

[0072] If |C xy |>ε (non-diagonal dominance), then V 1x _1=C xy V 1y _1= -C xx Otherwise (near diagonal matrix), V 1x _1=1, V 1y _1=0, and for V 1x _1 and V 1y Normalize _1 to obtain the eigenvector (V) corresponding to the largest eigenvalue. 1x V 1y For example, the normalization method can be Euclidean normalization (i.e., L2 normalization).

[0073] Furthermore, the formula for calculating the heading angle is θ1 = atan2(V 1x V 1y The heading angle represents the angle between the eigenvector corresponding to the largest eigenvalue and the radar longitudinal direction (i.e., the angle between the y-axis), and its value ranges from -180° to 180°.

[0074] For example, the heading angle of a target can be determined using a linear regression algorithm. The specific method is as follows:

[0075] The heading angle is determined by linearly fitting the position coordinates of multiple detection points and then using the straight line obtained from the linear fitting.

[0076] Linear fitting methods include least squares or Tukey M estimation, among others.

[0077] While linear regression is a relatively simple algorithm for determining heading angles, it suffers from poor performance for vertical point groups due to its inability to distinguish between independent and dependent variables. Furthermore, it is sensitive to outliers and lacks the robustness of heading angles calculated using principal component analysis. This method is suitable for points that are less distributed laterally but more distributed longitudinally.

[0078] For example, the target's heading angle can also be determined based on the principal axes of the convex hull. The specific method is as follows:

[0079] A point set is established based on the position coordinates of multiple detection points, and the convex hull of the point set is detected. Principal axis extraction is then performed on the convex hull to determine the target's heading angle. Principal axis extraction refers to extracting the principal axis from the minimum area bounding rectangle of the convex hull.

[0080] While using the principal axis of the convex hull to calculate the heading angle can provide stronger robustness to outliers, it has higher computational complexity and requires a higher number of detection points, with at least three non-collinear points needed.

[0081] For example, the target's heading angle can also be determined by linearly fitting the position coordinates of multiple detection points based on the Hall transform (Hough). The specific method is as follows:

[0082] The detection points are transformed into the Hough space, and voting is performed in the (ρ, θ) parameter space. The θ corresponding to the peak value is the heading angle.

[0083] Although this method is robust to outliers and missing data, it has high computational complexity and is not suitable for embedded real-time systems. Careful selection of angular resolution and accumulator threshold is required.

[0084] For example, the heading angle of a target can be determined based on a weighted principal component analysis algorithm. The specific method is as follows:

[0085] The detection points are weighted based on their signal-to-noise ratio (SNR). The higher the SNR, the greater the weight. A weighted covariance matrix is ​​calculated based on the weight of each detection point and its position coordinates. The heading angle is then calculated using principal component analysis based on the weighted covariance matrix.

[0086] The weighted covariance matrix can be expressed as:

[0087] Among them, W i This represents the weight of the i-th detection point. For the weighted mean, P i Let represent the original covariance matrix of the i-th detection point.

[0088] It is understandable that in this application, a two-dimensional point set can be constructed first based on the position coordinates of multiple detection points, and then the heading angle can be calculated. Alternatively, the heading angle can be calculated directly using the position coordinates of multiple detection points.

[0089] In step S03, the radial velocity of the target is determined based on the signal-to-noise ratio of multiple detection points.

[0090] For example, the method for obtaining the radial velocity of a target includes: obtaining the signal-to-noise ratio (SNR) of multiple detection points and comparing the SNR of the multiple detection points; and selecting the radial velocity of the detection point with the largest SNR as the radial velocity V_r of the target.

[0091] In non-rigid target scenarios, although the velocity directions of different parts are basically the same, the radial velocity amplitudes vary greatly. The radial velocity corresponding to the detection point with the highest signal-to-noise ratio is selected as the radial velocity of the target because a high signal-to-noise ratio usually corresponds to a strong scattering part of the target (such as the motorcycle body), and the radial velocity of this part is closest to the target's true motion speed.

[0092] For example, the method for obtaining the radial velocity of a target includes: obtaining the signal-to-noise ratio of multiple detection points, determining the weight of the radial velocity of each detection point based on the signal-to-noise ratio of the multiple detection points, and performing a weighted average of the radial velocities of the multiple detection points to obtain the radial velocity of the target, wherein the weight of the radial velocity of each detection point is positively correlated with its signal-to-noise ratio.

[0093] For example, the average value of the target's historical frame velocity can be used as the current velocity magnitude, and the radial velocity of the target can be obtained by balancing the velocity estimate through time filtering.

[0094] It is understood that in other embodiments of this application, the radial velocity of the target may also be the median or average of the radial velocities of multiple detection points.

[0095] In step S04, the target's velocity vector is calculated based on the target's heading angle and radial velocity.

[0096] Due to the eigenvector (V) 1x V 1y It can only determine the direction axis (a straight line passing through the origin), without distinguishing between positive and negative directions. Therefore, the calculation of the target velocity vector includes determining the characteristic vector (V) based on the target's relative motion direction, trajectory status, etc. 1x V 1yThe system checks whether the direction of the heading angle is the same as the actual direction of the target's movement. If the result is yes, a trigonometric function is used to synthesize the heading angle and the target's radial velocity into a velocity vector. If the result is no, a trigonometric function is used to synthesize the heading angle and the radial velocity after a 180° flip into a velocity vector. This ensures that the target's direction of movement always faces the front half of the radar coordinate system, which conforms to the physical constraint in actual traffic scenarios that the target's radial velocity is either positive (moving towards the radar) or negative (moving away from the radar).

[0097] The formula for calculating the velocity vector is: The method for calculating the velocity vector is V. x =V_r*cos(θ), V y =V_r*sin(θ), where θ represents the target's heading angle and V_r represents the target's radial velocity.

[0098] Furthermore, the radar target trajectory velocity acquisition method provided in this application also includes filtering the velocity vector.

[0099] For example, Kalman filters, linear filters (such as α-β-γ filters), and other filters can be used to filter the velocity vector.

[0100] When using a Kalman filter to filter the velocity vector, it is also necessary to initialize the state error covariance matrix of the Kalman filter. The initial error variance of the lateral velocity component in the state error covariance matrix is ​​greater than that of the longitudinal velocity component to reflect the higher uncertainty of the lateral velocity, thus providing a reasonable initial confidence boundary for the Kalman filter and enabling it to converge quickly. For example, the initial error variance of the lateral velocity component is 25.0, and the initial error variance of the longitudinal velocity component is 10.0.

[0101] Experiments show that for non-rigid targets, at the edge of the radar's field of view, the aforementioned radar track velocity acquisition method improves the trackhead velocity initialization success rate by over 50%. Furthermore, at the edge of the field of view where the azimuth error at the detection point is 5-10° and the radial velocity standard deviation is 3-5 m / s, the median error of the target heading angle is <10°, significantly better than existing technologies. In addition, the overall computation time is only 0.1 ms longer than existing technologies.

[0102] Figure 2 A flowchart illustrating another radar target track velocity acquisition method according to an embodiment of this application is shown.

[0103] See Figure 2 The radar target trajectory speed acquisition method provided in this application includes steps S11-S16.

[0104] In step S11, when the target's track head state is updated, the azimuth and radial velocity of multiple detection points related to the target's track are acquired, and the first track velocity acquisition method is used to acquire the target's first velocity vector.

[0105] The first track velocity acquisition method is the RANSAC angle-radial velocity fitting method. It is understood that the first track velocity acquisition method is not limited to this and can also be other existing radar track velocity acquisition methods.

[0106] In step S12, it is determined whether the first velocity vector is valid according to the preset conditions. If the determination result is negative, proceed to step S13, and if the determination result is positive, proceed to step S15.

[0107] Among them, the preset conditions include meeting the quality indicators such as the number of interior points, the proportion of interior points, and the angular span.

[0108] In step S13, it is determined whether the track age has reached the maximum track head threshold. If the determination result is yes, proceed to step S14.

[0109] If the result of this step is negative, the target's velocity vector may not be updated, or the target's velocity vector may be updated to the first velocity vector.

[0110] The maximum trackhead threshold can be set according to the actual situation, such as 3 frames or 5 frames.

[0111] In step S14, when the target is at the edge of the radar field of view, the second velocity vector of the target is obtained by the second trajectory velocity acquisition method, and when the target is not at the edge of the radar field of view, the third velocity vector of the target is obtained by the third trajectory velocity acquisition method.

[0112] The second trajectory speed acquisition method adopts, for example, Figure 1 The radar target trajectory velocity acquisition method, specifically the third trajectory velocity acquisition method, employs a trajectory velocity acquisition method that assumes the lateral motion velocity is 0.

[0113] It is understandable that the third method for obtaining track speed is not limited to this, and can also be used for other existing radar track speed acquisition methods.

[0114] In step S15, the target's velocity vector is updated.

[0115] The rules for updating the target's velocity vector are as follows: when the first velocity vector is valid, the first velocity vector is updated to the target's velocity vector; when the first velocity vector is invalid, if there is a second or third velocity vector, the second or third velocity vector is updated to the target's velocity vector; otherwise, the target's velocity vector is not updated.

[0116] Furthermore, in other embodiments of this application, step S14 may only include obtaining the second velocity vector of the target using the second track velocity acquisition method when the target is at the edge of the radar field of view.

[0117] Furthermore, in other embodiments of this application, when the first velocity vector is invalid, the second velocity vector of the target can also be calculated using the second trajectory velocity acquisition method when the track age reaches the maximum track head threshold or the target is at the edge of the radar field of view.

[0118] In step S16, the velocity vector of the target is filtered.

[0119] For example, a Kalman filter can be used to filter the velocity vector, and the state error covariance matrix of the Kalman filter is initialized before filtering. The initial error variance of the lateral velocity component in the state error covariance matrix is ​​greater than the initial error variance of the longitudinal velocity component.

[0120] Figure 3 This diagram illustrates how the principal component analysis algorithm is used to obtain the target heading angle in an embodiment of this application.

[0121] See Figure 3 When a target moves in a certain direction, the radar detects the scattering points from different parts of the target. Since the target has a certain geometric size in space, these scattering points are not concentrated at one point in the radar coordinate system, but rather spread linearly in the opposite direction of the movement.

[0122] For example, when a motorcycle travels at an angle θ to the radar's longitudinal direction, the projected positions of its front, body, and rear in the radar coordinate system form a "chain of points" along the θ direction. Even if there are errors in the angle measurement of individual detection points and the radial velocity is highly discrete, the spatial distribution of these points still retains the "geometric traces" of the direction of motion.

[0123] Since the eigenvector corresponding to the largest eigenvalue of the covariance matrix points in the direction of the largest data variance, for the point group distributed along the θ direction, its variance in the θ direction is much greater than its variance in the vertical direction. Therefore, the principal component direction θ naturally indicates the heading angle of the target.

[0124] Furthermore, this application also provides a radar target trajectory velocity initialization device, including a processor and a memory, wherein the memory is used to store computer instructions, and the processor executes the computer instructions to implement the radar target trajectory velocity acquisition method as described above.

[0125] Furthermore, this application also provides a storage medium storing a computer program that, when executed by a processor, implements the radar target track speed acquisition method as described above.

[0126] The radar target trajectory velocity acquisition method, apparatus, and storage medium provided in this application use the radial velocity of the detection point with the highest signal-to-noise ratio as the radial velocity of the target, or use the weighted average of the radial velocities of multiple detection points (the weight of the detection point is positively correlated with its signal-to-noise ratio) as the radial velocity of the target. This makes the radial velocity used for radar target trajectory velocity initialization closer to the actual radial velocity of the non-rigid target, thereby avoiding the interference of inconsistent radial velocities of multiple parts of the non-rigid target on the direction estimation and improving the reliability of the obtained velocity vector.

[0127] Furthermore, the target's heading angle is calculated using a robust principal component analysis (PCA) algorithm, ensuring that the velocity vector obtained from this heading angle balances robustness and reliability. This improves the effectiveness of initializing the trajectory velocity of non-rigid targets (e.g., obtaining an effective velocity vector even when the target is moving laterally or obliquely). In addition, PCA requires fewer iterations, has no requirements on the angular span, and theoretically only requires a minimum of two detection points to calculate the covariance matrix. Therefore, it can operate normally even in radar field-of-view edge scenarios with only 3-4 detection points.

[0128] Furthermore, different trajectory speed acquisition methods are selected according to different scenarios to calculate the target velocity vector, so that the optimal velocity vector of the target can be obtained in different application scenarios, achieving full scenario coverage.

[0129] As described above, these embodiments of the present application do not exhaustively describe all details, nor do they limit the invention to specific embodiments. Clearly, many modifications and variations can be made based on the above description. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the present application, thereby enabling those skilled in the art to make good use of the present application and modifications based on it. The scope of protection of this application should be determined by the scope defined by the claims of the present application and their equivalents.

Claims

1. A method for obtaining the velocity of a radar target track, comprising: Obtain the position coordinates of multiple detection points related to the target trajectory; The target's heading angle is determined based on the position coordinates of the multiple detection points; The radial velocity of the target is determined based on the signal-to-noise ratio of the multiple detection points; as well as The target's velocity vector is obtained based on the heading angle and the radial velocity.

2. The method of claim 1, wherein, The determination of the target's heading angle based on the position coordinates of multiple detection points includes: Calculate the covariance matrix based on the position coordinates of the multiple detection points; Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvector corresponding to the largest eigenvalue; and The heading angle is determined based on the eigenvector.

3. The method of claim 2, wherein, The velocity vector of the target obtained based on the heading angle and the radial velocity includes: Based on the target's relative motion direction and trajectory status, determine whether the direction of the feature vector is the same as the target's actual motion direction; When the judgment result is yes, the velocity vector is synthesized from the heading angle and the radial velocity using trigonometric functions; and If the result is negative, trigonometric functions are used to synthesize the heading angle and the radial velocity after being flipped 180° into the velocity vector.

4. The method of claim 1, wherein, Determining the radial velocity of the target based on the signal-to-noise ratio of the multiple detection points includes: Obtain the signal-to-noise ratio (SNR) of the multiple detection points and compare the SNR of the multiple detection points; and The radial velocity of the detection point with the highest signal-to-noise ratio is selected as the radial velocity of the target.

5. The method according to claim 1, wherein, Determining the radial velocity of the target based on the signal-to-noise ratio of the multiple detection points includes: Obtain the signal-to-noise ratio (SNR) of the multiple detection points, and determine the weight of the radial velocity at each detection point based on the SNR; and The radial velocity of the target is obtained by weighted averaging the radial velocities of the multiple detection points. The weight of the radial velocity at each detection point is positively correlated with its signal-to-noise ratio.

6. The method according to claim 4 or 5, further comprising: The velocity vector is filtered using a Kalman filter. Before filtering, the state error covariance matrix of the Kalman filter needs to be initialized so that the initial error variance of the lateral velocity component is greater than that of the longitudinal velocity component.

7. The method of claim 1, wherein, Determining the target's heading angle based on the position coordinates of the multiple detection points includes: The heading angle is determined by linearly fitting the position coordinates of the multiple detection points based on linear regression or Hall transform; or A weighted covariance matrix is ​​obtained based on the position coordinates of the multiple detection points, and the heading angle is determined using principal component analysis based on the weighted covariance matrix; or A point set is established based on the position coordinates of the multiple detection points, and the convex hull of the point set is detected. The principal axis of the convex hull is extracted to determine the heading angle.

8. A method for obtaining the velocity of a radar target track, comprising: When the target's track head status is updated, the azimuth and radial velocity of multiple detection points related to the target's track are acquired, and the first track velocity acquisition method is used to obtain the target's first velocity vector; When the first velocity vector is valid, the first velocity vector is updated to the target's velocity vector; as well as When the first velocity vector is invalid, and the track age reaches the maximum trackhead threshold and / or the target is at the edge of the radar field of view, the second track velocity acquisition method is used to obtain the target's second velocity vector, and the second velocity vector is updated to the target's velocity vector. The second trajectory speed acquisition method employs the method described in any one of claims 1-7.

9. The method according to claim 8, further comprising: When the first velocity vector is invalid, the track age reaches the maximum track head threshold, and the target is not at the edge of the radar field of view, the third track velocity acquisition method is used to obtain the target's third velocity vector, and the third velocity vector is updated to the target's velocity vector.

10. A radar target trajectory velocity initialization device, comprising: Memory, used to store computer instructions; The processor, when executing the computer instructions, implements the radar target track velocity acquisition method as described in any one of claims 1-9.

11. A storage medium storing a computer program, which, when executed by a processor, implements the radar target track velocity acquisition method as described in any one of claims 1 to 9.

Images