Target tracking method based on target course angle

Abstract

The invention relates to the technical field of vehicle-mounted millimeter wave radar tracking, in particular to a target tracking method based on a target course angle, and the method comprises the following steps: obtaining the position data of a plurality of targets around a vehicle; the multiple pieces of target position data are preprocessed; clustering the preprocessed data so as to position a target; and a track is established for the clustered target points, through multi-time frame accumulation, an improved RANSAC algorithm is utilized to fit the optimal motion track of the target, and the course angle of the target is estimated. Through the improved RANSAC algorithm and by utilizing the ideas of randomness and hypothesis, the method aims at solving the problem of calculation of the course angle of the target at the initial stage of the track target, and further improves the stability of target tracking and the accuracy of target speed calculation.

Description

Technical Field

[0001] The present application relates to the field of vehicle-mounted millimeter-wave radar tracking technology, and in particular to a target tracking method based on a target heading angle. Background Art

[0002] With the development of autonomous driving technology, the entire intelligent driving system has higher and higher requirements for the stability of target tracking and speed sensitivity. The on-board millimeter-wave radar is widely used in the entire system of autonomous driving due to its long detection distance, strong anti-interference ability, all-weather working ability and relatively accurate speed measurement ability. Since neither the visual system nor the lidar system can directly obtain the speed measurement of the target, the speed measurement capability of the millimeter-wave radar is crucial, and the speed estimation of the millimeter wave is closely related to the target heading angle estimation. Therefore, ensuring that the millimeter-wave radar accurately estimates the target heading angle is crucial for the safe driving of the vehicle. However, the existing target heading angle calculation method is not universal in certain specific scenarios and under different radars, and has high requirements for radar hardware, which limits its application in on-board millimeter-wave radar tracking technology.

[0003] During vehicle driving, due to the randomness of the target's motion direction, the target's heading angle cannot be accurately estimated by calculating a single frame alone, and the estimation of the target's starting heading angle directly leads to a slow convergence of the subsequent filtering process, or direct filtering divergence, resulting in target tracking failure. Therefore, the estimation of the target heading angle is crucial in the target's starting stage. In existing engineering applications, it is usually assumed that the target vehicle moves in a vertical direction, or the least squares method is used to fit the target's starting heading angle. Due to the influence of angle measurement errors and multipath reflections, such estimates usually lead to large errors, especially when the target moves quickly or in complex road environments, such as underground parking garages, traffic light intersections and other complex urban road conditions. Therefore, the problem of how to more accurately estimate the target's starting heading angle needs to be solved urgently. Summary of the invention

[0004] The technical problem to be solved by the present invention is that the existing automatic driving target tracking method has low accuracy.

[0005] To this end, the present invention provides a target tracking method based on target heading angle, which aims to solve the problem of calculating the target heading angle at the starting stage of the track target by using an improved RANSAC algorithm and its randomness and hypothetical ideas, and further improve the stability of target tracking and the accuracy of target speed calculation.

[0006] The technical solution adopted by the present invention to solve its technical problem is:

[0007] A target tracking method based on target heading angle comprises the following steps:

[0008] Step 1: Obtain the data of multiple target positions around the vehicle;

[0009] Step 2: Preprocess the data of multiple target positions;

[0010] Step 3: Cluster the preprocessed data to locate the target;

[0011] Step 4: Establish a track for the clustered target points. Through multi-time frame accumulation, and then use the improved RANSAC algorithm to fit the best motion trajectory of the target and estimate the target heading angle.

[0012] Further, in the above Step 2, the preprocessing of the target position data includes outlier rejection and coordinate transformation.

[0013] Further, the outlier rejection includes: calculating the average echo energy of all targets Setting a threshold parameter Rejecting the targets with energy less than of the set value.

[0014] Further, in the above Step 3, when clustering the preprocessed data, the echo points belonging to the same target are clustered into one class. According to the clustered echo points, a vehicle rectangular frame is simulated, and the center point of the rectangular frame is simulated as the equivalent point of the target measurement.

[0015] Further, in Step 4, select the equivalent point of the nth target at the tth time frame as Pnt, associate this point with the temporary track of this target at the (t - 1)th time frame. After successful association, use Pnt as the selected point for the start of the track. If Pnt cannot be associated with the temporary track of this target at the (t - 1)th time frame, then this point becomes a new temporary track.

[0016] Further, when the number of equivalent points in a certain temporary track is greater than the threshold N, upgrade this temporary track to a stable track, that is, the number of equivalent points in the stable track is N + 1. Arbitrarily select two equivalent points in the stable track to fit a straight line, calculate the distance between other equivalent points in this stable track and this fitted straight line, and calculate the mean value Ed of all distances i1 and variance All pairs of equivalent points in a stable track form fitted straight lines.

[0017] Further, select the combination with the smallest mean value and the fitted straight line Ax + By + C = 0 with the smallest variance to approximately estimate the target heading angle

[0018] Further, if the fitted straight line with the smallest mean value and the fitted straight line with the smallest variance are the same one, then this fitted straight line can be approximated as the target motion trajectory.

[0019] Further, if there are two fitting lines with the minimum mean and two fitting lines with the minimum variance respectively, compare whether the mean of the combination with the minimum variance is less than the set mean threshold ε E , if it is less than the threshold, select the combination with the minimum variance; if it is greater than the threshold, select the combination with the minimum mean.

[0020] The beneficial effect of the present invention is that this application utilizes the robustness of the RANSAC algorithm. By using the randomness and hypothesis characteristics of this algorithm, it can effectively solve the stability and accuracy of radar target heading angle estimation in complex road environments.

[0021] This application uses an improved RANSAC algorithm to estimate the target starting heading angle, and can effectively estimate relatively accurate parameters from data containing outliers. The core lies in randomly selecting sample points in the dataset for model fitting and iterating this process to find the best model, effectively improving the accuracy of target heading angle estimation in complex road environments.

[0022] By effectively enhancing the robustness of target heading angle estimation, as described above, based on the target heading angle, the true motion direction and target speed of the target can be estimated relatively accurately, providing a basis for the subsequent filtering process, thereby improving the stability and accuracy of radar multi-target tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described below in conjunction with the drawings and embodiments.

[0024] Figure 1 is a schematic structural diagram of the present invention.

[0025] Figure 2 is a schematic diagram of radar measurement in the present invention.

[0026] Figure 3 is a schematic diagram of target position data clustering in the present invention.

[0027] Figure 4 is a schematic diagram of RANSAC fitting in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0029] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. are based on the orientation or positional relationships shown in the drawings. These are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention. In addition, features defined as "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more.

[0030] In the description of the present invention, it should be noted that unless otherwise clearly specified and defined, the terms "mounted", "connected", and "coupled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0031] A target tracking method based on a target heading angle includes the following steps:

[0032] Step 1, obtaining targets around the vehicle through a millimeter-wave radar

[0033] Millimeter-wave radars are respectively installed at the left and right bumpers in front of and behind the vehicle, and a millimeter-wave radar is installed directly in front of the vehicle. The position data of the targets around the vehicle is obtained in real time through the millimeter-wave radars.

[0034] Step 2, data preprocessing

[0035] The raw data collected by the radar system is preprocessed. Since the collected raw data contains noise and interference, the preprocessing process mainly uses signal processing techniques to filter out the noise and interference in the raw data according to the echo energy of the target and the RCS value of the target. The main steps include:

[0036] S2.1 Outlier rejection

[0037] Calculate the average value of the echo energy of all targets Set the threshold parameter Reject targets with energy less than ;

[0038] S2.2 Coordinate transformation

[0039] The radial distance of the target measured by the radar is R, and the azimuth angle of the target is α. As Figure 1 shown, the horizontal and vertical distances of the target reflection point can be calculated as follows:

[0040] Y = R * sinα

[0041] X = R * cosα

[0042] Step 3: Target positioning

[0043] Referring to Figure 4 , vehicle targets on the road are generally extended targets, and the echo signals are generally multiple points. Therefore, it is necessary to cluster the preprocessed data, cluster the echo points belonging to the same target into one class. This step can use the DBSCAN algorithm for clustering. According to the clustered echo points, a vehicle rectangle is simulated, and the center point of the rectangle is simulated as the equivalent point of the target measurement, providing a basis for subsequent track association and heading angle calculation.

[0044] Step 4: Establish a track for the clustered target points. Through multi-frame accumulation, and then use the improved RANSAC algorithm to fit the best motion track of the target, so as to estimate the heading angle of the target. The improved RANSAC algorithm is set in the Heading Angle Estimate Module (HAEM). Specifically, its calculation steps include:

[0045] S4.1 Select a preprocessed equivalent point as the selected point for the start of the track. Through a suitable association algorithm (such as the nearest neighbor / Hungarian algorithm, etc.), establish an association relationship between this point and the temporary track of this target, and record the position information of this point, establish a temporary track for maintenance. The associated reflection point is no longer used as the selected point for the start of the track. Cycle through all the equivalent points in the current period and repeat the above process. If the temporary track fails to establish an association relationship with a new equivalent point for 3 consecutive periods, it is determined that the temporary track terminates, and this track is deleted from the track queue.

[0046] Specifically, assume that the vehicle identifies n targets (n = 1, 2, 3...). Define the equivalent point of the nth target in the tth time frame as P nt , select P nt as the selected point for the start of the track, associate this point with the temporary track of this target in the (t - 1)th time frame of this target. After association, use P nt as the selected point for the start of the track;

[0047] If P nt cannot be associated with the temporary track of this target in the (t - 1)th time frame, then this point becomes a new temporary track.

[0048] S4.2 When the number of equivalent points in a temporary track is greater than the threshold N, upgrade the temporary track to a stable track. The stable track is the set of equivalent points, and the number of equivalent points in the stable track set is N + 1.

[0049] Estimate the heading angle of the stable track using the improved RANSAC algorithm. Randomly select two equivalent points p1 and p2 in the set, as Figure 4 shown by the two square points in the figure. Fit the two selected points into a straight line Ax + By + C = 0. Other equivalent points not on the line are called outliers, as Figure 4 shown by the blue points.

[0050] Calculate the distance from all outliers (x i , y i ) to the fitted line:

[0051]

[0052] And calculate the mean Ed i1 and variance of all distances:

[0053]

[0054] S4.3 Continue to randomly select two non-repeating points in the set and repeat step S4.2 until all pairwise combinations in the data set have been cycled through. Finally, the means and variances of all combinations can be obtained.

[0055] S4.4 Select the combination with the minimum mean and the combination with the minimum variance respectively. If they are the same combination, the line fitted by this combination can be approximately regarded as the target motion trajectory; if they are not the same combination, compare whether the mean of the combination with the minimum variance is less than the set mean threshold ε E (This threshold can be adjusted according to different radar detection angle errors and tracking data experience). If it is less than the threshold, select the combination with the minimum variance; if it is greater than the threshold, select the combination with the minimum mean.

[0056] Fit a line to this equivalent point combination to approximately estimate the heading angle of the target. A and B are the coefficients of this line.

[0057] Through the above calculations, the robustness of the target heading angle estimation can be effectively enhanced. As mentioned above, based on the target heading angle, the true motion direction and target speed of the target can be estimated relatively accurately, providing a basis for the subsequent filtering process, thereby improving the stability and accuracy of radar target tracking.

[0058] The traditional method of calculating outliers in the RANSAC algorithm is that the distance from a point to a line is greater than a certain threshold, and finally the final fitted curve is determined according to the number of outliers. However, at the initial stage of target tracking, in order to perform fast tracking, the number of valid points within the initial stable track of the target is usually no more than 5-6, which is a small number. The above calculation method is not suitable for this working condition. Therefore, in this patent, the mean and variance information of the distance from outliers to the line are compared to obtain the final combination of fitted lines.

[0059] Inspired by the ideal embodiments of the present invention described above, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of the present invention. The technical scope of the present invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A target tracking method based on a target heading angle, characterized in that, It includes the following steps: Step 1: Obtain the data of multiple target positions around the vehicle; Step 2: Preprocess the data of multiple target positions; Step 3: Cluster the preprocessed data to locate the target; Step 4: Establish a track for the clustered target points. Through multi-time frame accumulation, and then use the improved RANSAC algorithm to fit the best motion trajectory of the target and estimate the target heading angle.

2. The target tracking method based on the target heading angle according to claim 1, wherein In Step 2, the preprocessing of the target position data includes outlier rejection and left-side conversion.

3. The target tracking method based on the target heading angle according to claim 2, wherein The wild value rejection includes: calculating the average echo energy of all targets Setting a threshold parameter Rejecting targets with energy less than thereof 4. The target tracking method based on a target heading angle according to claim 1, wherein In Step 3, cluster the preprocessed data, cluster the echo points belonging to the same target into one category. According to the clustered echo points, simulate a vehicle rectangular frame, and simulate the center point of the rectangular frame as the equivalent point of the target measurement.

5. The target tracking method based on a target heading angle according to claim 4, wherein In step four, select the equivalent point of the nth target at the tth time frame as P nt , associate this point with the temporary track of this target at the (t - 1)th time frame. After successful association, use P nt as the selected point for the start of the track. If P nt cannot be associated with the temporary track of this target at the (t - 1)th time frame, then this point becomes a new temporary track.

6. The target tracking method based on a target heading angle according to claim 5, characterized in that When the number of equivalent points in a temporary track is greater than the threshold N, upgrade this temporary track to a stable track, that is, the number of equivalent points in the stable track is N + 1. Arbitrarily select two equivalent points in the stable track to fit a straight line, calculate the distance between other equivalent points in the stable track and this fitted straight line, and calculate the mean value Ed of all distances i1 Sum of squares of deviations All pairs of equivalent points in a stable track are combined to obtain fitted straight lines 7. The target tracking method based on a target heading angle according to claim 6, characterized in that, Select the combination with the minimum mean and the fitted straight line Ax + By + C = 0 with the minimum variance to approximately estimate the heading angle of the target 8. The target tracking method based on a target heading angle according to claim 7, wherein If the fitting line with the minimum mean and the fitting line with the minimum variance are the same, then this fitting line can be approximated as the target motion trajectory.

9. The target tracking method based on a target heading angle according to claim 7, wherein If there are two fitting lines with the minimum mean and the minimum variance respectively, compare whether the mean of the combination with the minimum variance is less than the set mean threshold ε E If it is less than the threshold, select the combination with the minimum variance; if it is greater than the threshold, select the combination with the minimum mean.

Images