A target tracking method, device and communication equipment

Abstract

The present invention provides a target tracking method, apparatus and communication device. The method includes: performing target perception by using at least two 4D millimeter-wave radars, each 4D millimeter-wave radar calculating target speed information, clustering the collected point cloud to obtain a target point cloud cluster and the central position of the target point cloud cluster; calculating a target heading angle and a target pitch angle based on the speed information of the same perceived target in multiple frames and the central position of the target point cloud cluster by using a stochastic parallel gradient descent algorithm; determining a target minimum bounding box according to the target heading angle and the target pitch angle, and obtaining the position information of the target. Based on the current situation of sparse point cloud of 4D millimeter-wave radar, the present invention performs upsampling processing on the point cloud collected by the 4D millimeter-wave radar, determines the target minimum bounding box by using the target heading angle and the target pitch angle, and obtains the position information of the target, thereby realizing target tracking and improving the accuracy of target tracking and perception.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a target tracking method, device and communication device. Background Art

[0002] Most of the existing millimeter-wave radars are 3D millimeter-wave radars, which refer to the detection of target positions and target speed information on a 2D plane. The target detection based on 4D millimeter-wave radars mainly obtains the target point cloud clusters through point cloud position and speed clustering, and then obtains the target position and speed information.

[0003] Due to the limited angular resolution of 4D millimeter-wave radars, effective calibration between devices cannot be performed during cross-station perception, resulting in ineffective matching of targets during cross-station tracking and inability to achieve target tracking. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a target tracking method, device and communication device to solve the technical problem that target tracking cannot be achieved by 4D millimeter-wave radars during cross-station perception in the prior art.

[0005] To solve the above problems, the embodiments of the present invention provide a target tracking method, which includes:

[0006] Perceiving a target by using at least two 4D millimeter-wave radars. Each 4D millimeter-wave radar calculates the target speed information, and clusters the collected point cloud to obtain the target point cloud cluster and the central position of the target point cloud cluster;

[0007] Based on the speed information of the same perceived target in multiple frames and the central position of the target point cloud cluster, calculate the target heading angle and target pitch angle based on the random parallel gradient descent algorithm;

[0008] Determine the target minimum bounding box according to the target heading angle and the target pitch angle to obtain the position information of the target.

[0009] Among them, the determining the target minimum bounding box according to the target heading angle and the target pitch angle includes:

[0010] Obtain a tracking target list and a perceived target list. The tracking target list includes target sampled point cloud, probability prediction value corresponding to the point cloud, target Kalman state variables, Doppler speed information and target ID. The perceived target list includes target measurement point cloud and Doppler speed information;

[0011] Calculate the association between the tracking target and the perceived target according to the Hungarian algorithm to obtain the matching pair of the tracking target and the perceived target;

[0012] Calculate the position of the upsampled point of the tracking target according to the upsampling historical position and Doppler velocity information;

[0013] Calculate the position credibility and status credibility of the upsampled point according to the position of the upsampled point of the tracking target;

[0014] Determine the minimum bounding box of the target according to the position credibility and status credibility of the upsampled point;

[0015] Among them, calculating the position of the upsampled point of the tracking target according to the upsampling historical position and Doppler velocity information includes:

[0016] Calculate the position of the upsampled point of the tracking target according to the historical coordinates of the upsampled point cloud, the time difference between the historical data time and the current time, and the Doppler velocity at the historical time;

[0017]

[0018] Among them, x, y, z are the historical coordinates of the upsampled point cloud, x′, y′, z′ are the coordinates of the upsampled point of the tracking target, ΔT is the time difference between the historical data time and the current time, and v x , v y , v z is the Doppler velocity recorded at the historical time.

[0019] Among them, calculating the position credibility and status credibility of the upsampled point includes:

[0020] Construct the perceived target box and virtual target box according to the tracking target list and the perceived target list;

[0021] Determine the credibility and status credibility of the tracking target according to the relationship between the tracking target and the perceived target box and the virtual target box.

[0022] Among them, determining the credibility and status credibility of the tracking target according to the relationship between the tracking target and the perceived target box and the virtual target box includes:

[0023] When the tracking target is within the perceived target box, the credibility of the tracking target is 1, and the status credibility of the tracking target is the sum of the default value and 0.5;

[0024] When the tracking target is within the virtual target box and not within the perceived target box, the credibility of the tracking target is P = d * lp / (8 * l l * l w * l h*ld), the state confidence of the tracking target is δ = δ - n + P, where P is the confidence of the tracking target, d is the Euclidean distance from the upsampled point to the center point of the virtual box, lp is the distance from the tracking target to the outside of the perceived target box, ld is the distance from the tracking target to the inside of the virtual target box, and l l ,l w ,l h is the size of the virtual target box, δ is the state confidence of the tracking target, and n is the frame difference between the perceived target and the tracking target;

[0025] When the tracking target is outside the virtual target box, the confidence of the tracking target is

[0026] the state confidence of the tracking target is δ = δ - n, lo is the distance from the tracking target to the virtual target box, and lc is the distance from the tracking target to the perceived target box.

[0027] Among them, determining the target minimum bounding box according to the position confidence and state confidence of the upsampled point includes:

[0028] Calculating the first position confidence of the upsampled point in the perceived target point cloud cluster;

[0029] Calculating the second position confidence of the upsampled point in the tracking target point cloud cluster;

[0030] When the first position confidence and the second position confidence meet the same conditions, the target minimum bounding box is determined.

[0031] Among them, the 4D millimeter-wave radar is a radar-vision integrated machine.

[0032] An embodiment of the present invention further provides a target tracking device, which includes a processing module, a calculation module, and a target tracking module;

[0033] The processing module is used to perform target perception by using at least two 4D millimeter-wave radars. Each 4D millimeter-wave radar calculates target speed information, clusters the collected point cloud to obtain a target point cloud cluster and the center position of the target point cloud cluster;

[0034] The calculation module is used to calculate the target heading angle and target pitch angle based on the speed information of multiple frames of the same perceived target and the center position of the target point cloud cluster by using the random parallel gradient descent algorithm;

[0035] The target tracking module is used to determine the target minimum bounding box according to the target heading angle and the target pitch angle obtained by the calculation module, and obtain the position information of the target.

[0036] Among them, the target tracking module includes:

[0037] An acquisition unit, configured to acquire a tracking target list and a sensed target list, where the tracking target list includes target sampling point clouds and probability prediction values corresponding to the point clouds, target Kalman state variables, Doppler velocity information, and target IDs, and the sensed target list includes target measurement point clouds and Doppler velocity information;

[0038] An association unit, configured to calculate the association between the tracking targets and the sensed targets according to the Hungarian algorithm to obtain the matching pairs of the tracking targets and the sensed targets;

[0039] A first calculation unit, configured to calculate the positions of the upsampled points of the tracking targets according to the upsampled historical positions and the Doppler velocity information;

[0040] A second calculation unit, configured to calculate the position credibility and state credibility of the upsampled points according to the positions of the upsampled points of the tracking targets;

[0041] A target tracking unit, configured to determine the minimum bounding box of the target according to the position credibility and state credibility of the upsampled points.

[0042] Among them, calculating the position credibility and state credibility of the upsampled points includes:

[0043] Constructing sensed target boxes and virtual target boxes according to the tracking target list and the sensed target list;

[0044] Determining the credibility and state credibility of the tracking targets according to the relationships between the tracking targets, the sensed target boxes, and the virtual target boxes.

[0045] An embodiment of the present invention further provides a communication device, including a memory, a processor, and a program stored on the memory and executable on the processor. When the processor executes the program, the above-mentioned target tracking method is implemented.

[0046] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps in the above-mentioned target tracking method are implemented.

[0047] The above technical solutions of the present invention have at least the following beneficial effects:

[0048] In the target tracking method, gateway, and communication device provided by the embodiments of the present invention, at least two 4D millimeter-wave radars are used for target perception. Each 4D millimeter-wave radar calculates the target speed information, and based on the stochastic parallel gradient descent algorithm, the target heading angle and the target pitch angle are calculated. Then, according to the target heading angle and the target pitch angle, the minimum bounding box of the target is determined, and the position information of the target is obtained. Based on the current situation of sparse point clouds of 4D millimeter-wave radars, the embodiments of the present invention perform upsampling processing on the point clouds collected by the 4D millimeter-wave radars, determine the minimum bounding box of the target by using the target heading angle and the target pitch angle, and obtain the position information of the target, thereby realizing target tracking and improving the accuracy of target tracking and perception. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It shows the flowchart of the target tracking method provided by the embodiments of the present invention;

[0050] Figure 2 It shows the schematic flowchart of determining the target position information in the target tracking method provided by the embodiments of the present invention;

[0051] Figure 3 It shows the schematic structural diagram of the target tracking device provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] To make the technical problems, technical solutions, and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0053] As Figure 1 shown, the embodiments of the present invention provide a target tracking method, including:

[0054] Step 101: Use at least two 4D millimeter-wave radars for target perception. Each 4D millimeter-wave radar calculates the target speed information, and clusters the collected point clouds to obtain the target point cloud cluster and the central position of the target point cloud cluster;

[0055] The target tracking method of the embodiments of the present invention is applied to the road test perception scenario. Multiple 4D millimeter-wave radars are deployed for road tests, and the background point clouds of the 4D millimeter-wave radars within a certain period of time are extracted. For example, 30s of background point clouds are collected and aggregated into background point cloud information; the 4D millimeter-wave radars of the embodiments of the present invention can be radar-vision integrated machines;

[0056] Three millimeter-wave radar calibrators are placed 30 meters in front of each radar-vision integrated machine. Through the background point cloud information, the positions of the millimeter-wave radar calibrators in the 4D millimeter-wave radar coordinate system are extracted, and the world coordinates of the centers of the four millimeter-wave radars are calculated through high-precision positioning. Through the coordinate conversion relationship, the conversion relationship between the radar-vision integrated machine and the world coordinate system is calculated.

[0057] Step 102: Based on the speed information of multiple frames of the same perceived target and the center position of the target point cloud cluster, calculate the target heading angle and target pitch angle using the stochastic parallel gradient descent algorithm;

[0058] In the embodiment of the present invention, in order to obtain the accurate displacement between devices and the included angle of the perception coordinate system, considering the resolution and high-precision positioning error, which may lead to inaccurate calibration of the pitch angle and heading angle of the radar-vision integrated machine, an optimization of the external parameters is proposed.

[0059] For two adjacent radar-vision integrated machines that can both perceive the target within a certain area, each device calculates the target speed information through the Doppler effect. By synthesizing the speed information of multiple frames of the same perceived target and the vehicle center information, an optimization algorithm for the heading angle and pitch angle based on the stochastic parallel gradient descent algorithm is carried out.

[0060] First, determine the initial conversion matrix between 4D millimeter-wave radars. In this embodiment, the heading angle, pitch angle interpolation α, β between radars, and the coordinate differences Δx, Δy, Δz of the radar center positions are calculated through the initial calibration information, and the conversion relationship between the two millimeter-wave radar coordinate systems is obtained as follows:

[0061]

[0062] Among them,

[0063] Then, set the perturbation values for the perturbation variables. The perturbation variables are the heading angle and pitch angle, and the corresponding perturbation values are Δα and Δβ. The perturbation interval is set to ±0.1 times the angle accuracy of the millimeter-wave radar. Update the pitch angle and heading angle, and substitute them into the conversion relationship between the two millimeter-wave radar coordinate systems to form the target heading and center point of millimeter-wave radar 1.

[0064] Considering the influence of angle accuracy and angular resolution, the 4D millimeter-wave radar target similarity analysis function is as follows: It is considered that the target point cloud of millimeter-wave radar 1 is the rearmost point cloud.

[0065]

[0066] Where

[0067] In the formula, x1 and y1 are the horizontal and vertical coordinates of the corresponding points of millimeter-wave radar 1 in the matching pair, x2 and y2 are the horizontal and vertical coordinates of the corresponding points of millimeter-wave radar 2, s is the radar distance accuracy parameter, G is the Gaussian filtering function d = V1·V2.

[0068] Step 103: Determine the target minimum bounding box according to the target heading angle and target pitch angle to obtain the position information of the target.

[0069] In the embodiment of the present invention, the Kalman filtering algorithm is used for target tracking. Since the 4D millimeter-wave radar point cloud is sparse, the point cloud data is upsampled.

[0070] The implementation process of this step includes:

[0071] Step 1031: Obtain the tracking target list and the sensed target list. The tracking target list includes the target sampled point cloud and the probability prediction value corresponding to the point cloud, the target Kalman state variables, the Doppler velocity information, and the target ID. The sensed target list includes the target measurement point cloud and the Doppler velocity information;

[0072] Step 1032: Calculate the association between the tracking target and the sensed target according to the Hungarian algorithm to obtain the matching pairs of the tracking target and the sensed target;

[0073] Since the tracking target reflects the position of the target at the previous frame moment, by predicting the current position, it is associated with the sensed target on the current road, and it is determined that they are the same target. In this embodiment, the association between the tracking target and the sensed target is calculated according to the Hungarian algorithm to obtain the matching pairs of the tracking target and the sensed target;

[0074] Step 1033: Calculate the position of the upsampled point of the tracking target according to the upsampled historical position and the Doppler velocity information;

[0075] Specifically, according to the historical coordinates of the upsampled point cloud, the time difference between the historical data moment and the current moment, and the Doppler velocity at the historical moment, the position of the upsampled point of the tracking target is calculated

[0076]

[0077] where x, y, z are the historical coordinates of the upsampled point cloud, x′, y′, z′ are the coordinates of the upsampled point of the tracking target, ΔT is the time difference between the historical data moment and the current moment, and v x , v y , v z is the Doppler velocity recorded at the historical moment.

[0078] As a preferred embodiment, an error perturbation function is obtained by adding an error within the distance accuracy to each upsampled point as follows:

[0079]

[0080] where x n , y n , z n are the point coordinates after adding the perturbation, and Δx, Δy, Δz are the distance perturbations less than the distance resolution, which can be set to perturbations less than 0.1 m in the embodiment.

[0081] Step 1034: Calculate the position credibility and status credibility of the upsampled points based on the positions of the tracking targets.

[0082] As a preferred embodiment, the method for determining the credibility and status credibility of the tracking target includes: constructing a perceived target box and a virtual target box based on the tracking target list and the perceived target list; determining the credibility and status credibility of the tracking target according to the relationship between the tracking target and the perceived target box and the virtual target box.

[0083] Specifically, determining the credibility and status credibility of the tracking target according to the relationship between the tracking target and the perceived target box and the virtual target box includes:

[0084] When the tracking target is within the perceived target box, the credibility of the tracking target is 1, and the status credibility of the tracking target is the sum of the default value and 0.5.

[0085] When the tracking target is within the virtual target box and not within the perceived target box, the credibility of the tracking target is P = d * lp / (8 * l l *l w *l h *ld), and the status credibility of the tracking target is δ = δ - n + P, where P is the credibility of the tracking target, d is the Euclidean distance from the upsampled point to the center point of the virtual box, lp is the distance from the tracking target to the outside of the perceived target box, ld is the distance from the tracking target to the inside of the virtual target box l l ,l w ,l h is the size of the virtual target box, δ is the status credibility of the tracking target, and n is the frame difference between the perceived target and the tracking target.

[0086] When the tracking target is outside the virtual target box, the credibility of the tracking target is

[0087] The status credibility of the tracking target is δ = δ - n, where lo is the distance from the tracking target to the virtual target box, and lc is the distance from the tracking target to the perceived target box.

[0088] Step 1035: Determine the target minimum bounding box based on the position credibility and status credibility of the upsampled points.

[0089] Specifically, calculate the position credibility and status credibility of each virtual upsampled point, remove the points with status credibility less than 1, and form the final target minimum bounding box through the perceived target point cloud cluster and the points with P > 0.3 in the virtual upsampled points. Add the perceived target point cloud cluster to the virtual sampled points and set the initial status credibility of each point to 4.

[0090] As a preferred embodiment, determining a target minimum bounding box according to the position credibility and status credibility of upsampling points includes:

[0091] Calculating a first position credibility of upsampling points in the perceived target point cloud cluster;

[0092] Calculating a second position credibility of upsampling points in the tracked target point cloud cluster;

[0093] When the first position credibility and the second position credibility meet the same conditions, determining the target minimum bounding box.

[0094] The target tracking method of the embodiment of the present invention, based on the current situation of sparse point clouds of 4D millimeter-wave radar, performs upsampling processing on the point clouds collected by the 4D millimeter-wave radar, determines the target minimum bounding box by using the target heading angle and the target pitch angle, and obtains the position information of the target, thereby realizing target tracking and improving the accuracy of target tracking and perception.

[0095] The embodiment of the present invention provides a target tracking device, as Figure 3 shown. The device includes a processing module 301, a calculation module 302, and a target tracking module 303;

[0096] The processing module 301 is configured to perform target perception by using at least two 4D millimeter-wave radars. Each 4D millimeter-wave radar calculates target speed information, and clusters the collected point clouds to obtain a target point cloud cluster and the central position of the target point cloud cluster;

[0097] The calculation module 302 is configured to calculate a target heading angle and a target pitch angle based on the speed information of the same perceived target in multiple frames and the central position of the target point cloud cluster by using a stochastic parallel gradient descent algorithm;

[0098] The target tracking module 303 is configured to determine a target minimum bounding box according to the target heading angle and the target pitch angle obtained by the calculation module, and obtain the position information of the target.

[0099] The 4D millimeter-wave radar in the embodiment of the present invention is a radar-vision integrated machine.

[0100] As a preferred embodiment, the target tracking module includes:

[0101] An acquisition unit, configured to acquire a tracking target list and a perception target list. The tracking target list includes target sampled point clouds and probability prediction values corresponding to the point clouds, target Kalman state variables, Doppler speed information, and target IDs. The perception target list includes target measurement point clouds and Doppler speed information;

[0102] An association unit, configured to calculate the association between a tracking target and a perception target according to the Hungarian algorithm, and obtain a matching pair between the tracking target and the perception target;

[0103] A first calculation unit, configured to calculate the position of the upsampled point of the tracking target according to the upsampled historical position and Doppler velocity information;

[0104] A second calculation unit, configured to calculate the position credibility and state credibility of the upsampled point according to the position of the upsampled point of the tracking target;

[0105] A target tracking unit, configured to determine a target minimum bounding box according to the position credibility and state credibility of the upsampled point.

[0106] As a preferred embodiment, calculating the position credibility and state credibility of the upsampled point in the second calculation unit includes:

[0107] Constructing a perception target box and a virtual target box according to a tracking target list and a perception target list;

[0108] Determining the credibility and state credibility of the tracking target according to the relationship between the tracking target and the perception target box and the virtual target box.

[0109] As a preferred embodiment, the first calculation unit includes:

[0110] Calculating the position of the upsampled point of the tracking target according to the upsampled point cloud historical coordinates, the time difference between the historical data time and the current time, and the Doppler velocity at the historical time;

[0111]

[0112] Where x, y, z are the upsampled point cloud historical coordinates, x′, y′, z′ are the upsampled point coordinates of the tracking target, ΔT is the time difference between the historical data time and the current time, and v x , v y , v z Is the Doppler velocity recorded at the historical time.

[0113] As a preferred embodiment, determining the credibility and state credibility of the tracking target according to the relationship between the tracking target and the perception target box and the virtual target box in the second calculation unit includes:

[0114] When the tracking target is within the perception target box, the credibility of the tracking target is 1, and the state credibility of the tracking target is the sum of the default value and 0.5;

[0115] When the tracking target is within the virtual target box and not within the perception target box, the credibility of the tracking target is P = d * lp / (8 * ll *l w *l h *ld), the state confidence of the tracking target is δ = δ - n + P, where P is the confidence of the tracking target, d is the Euclidean distance from the upsampling point to the center point of the virtual box, lp is the distance from the tracking target to the outside of the perception target box, and ld is the distance from the tracking target to the inside of the virtual target box. The l l ,l w ,l h is the size of the virtual target box, δ is the state confidence of the tracking target, and n is the frame difference between the perception target and the tracking target;

[0116] When the tracking target is outside the virtual target box, the confidence of the tracking target is

[0117] The state confidence of the tracking target is δ = δ - n, lo is the distance from the tracking target to the virtual target box, and lc is the distance from the tracking target to the perception target box.

[0118] As a preferred embodiment, the target tracking unit includes: calculating the first position confidence of the upsampling points in the perception target point cloud cluster; calculating the second position confidence of the upsampling points in the tracking target point cloud cluster; when the first position confidence and the second position confidence meet the same conditions, determining the target minimum bounding box.

[0119] Based on the current situation of sparse 4D millimeter-wave radar point cloud, the target tracking device of the embodiment of the present invention performs upsampling processing on the point cloud collected by the 4D millimeter-wave radar, determines the target minimum bounding box by using the target heading angle and the target pitch angle, obtains the position information of the target, and thus realizes target tracking, which can improve the accuracy of target tracking and perception.

[0120] It should be noted that the communication device provided in the embodiment of the present invention is a communication device capable of executing the above target tracking method. Then, all embodiments of the above target tracking method are applicable to this communication device and can achieve the same or similar beneficial effects.

[0121] The embodiment of the present invention also provides a communication device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it realizes each process in the above target tracking method embodiment and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0122] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements each process in the above-mentioned embodiment of the target tracking method and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium includes, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0123] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0124] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one process Figure 1 or multiple processes and / or one block or multiple blocks.

[0125] These computer program instructions can also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable storage medium generate a paper product including an instruction device, and the instruction device implements the functions specified in one process Figure 1 or multiple processes and / or blocks Figure 1 or multiple blocks.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, so that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in one process Figure 1 or multiple processes and / or blocks Figure 1 or multiple blocks.

[0127] The above are the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A target tracking method, characterized in that, The method includes: Performing target perception using at least two 4D millimeter-wave radars. Each 4D millimeter-wave radar calculates target velocity information, clusters the collected point cloud to obtain a target point cloud cluster and the central position of the target point cloud cluster; Based on the velocity information of the same perceived target in multiple frames and the central position of the target point cloud cluster, calculating the target heading angle and target pitch angle based on the stochastic parallel gradient descent algorithm; Determining a target minimum bounding box according to the target heading angle and the target pitch angle to obtain the position information of the target; The determining the target minimum bounding box according to the target heading angle and the target pitch angle includes: Obtaining a tracking target list and a perceived target list. The tracking target list includes target sampled point cloud and the probability prediction value corresponding to the point cloud, target Kalman state variables, Doppler velocity information, and target ID. The perceived target list includes target measurement point cloud and Doppler velocity information; Calculating the association between the tracking target and the perceived target according to the Hungarian algorithm to obtain the matching pairs of the tracking target and the perceived target; Calculating the position of the upsampled point of the tracking target according to the upsampled historical position and Doppler velocity information; Calculating the position confidence and state confidence of the upsampled point according to the position of the upsampled point of the tracking target; Determining the target minimum bounding box according to the position confidence and state confidence of the upsampled point.

2. The method according to claim 1, wherein Calculating the position of the upsampled point of the tracking target according to the upsampled historical position and Doppler velocity information includes: Calculating the position of the upsampled point of the tracking target according to the historical coordinates of the upsampled point cloud, the time difference between the historical data time and the current time, and the Doppler velocity at the historical time; Among them, x, y, z are the historical coordinates of the upsampled point cloud, x′, y′, z′ are the coordinates of the upsampled points of the tracking target, ΔT is the time difference between the historical data time and the current time, v x , v y , v z is the Doppler velocity recorded at the historical time.

3. The method according to claim 1, characterized in that, The calculating the position confidence and state confidence of the upsampled point includes: Constructing a perceived target box and a virtual target box according to the tracking target list and the perceived target list; Determining the confidence and state confidence of the tracking target according to the relationship between the tracking target and the perceived target box and the virtual target box.

4. The method according to claim 3, characterized in that Determining the confidence and state confidence of the tracking target according to the relationship between the tracking target and the perceived target box and the virtual target box includes: When the tracking target is within the perceived target box, the confidence of the tracking target is 1, and the state confidence of the tracking target is the sum of the default value and 0.5; When the tracking target is within the virtual target box but not within the perceived target box, the confidence level of the tracking target is P = d * lp / (8 * l l *l w *l h *ld), and the state confidence level of the tracking target is δ = δ - n + P, where P is the confidence level of the tracking target, d is the Euclidean distance from the upsampling point to the center point of the virtual target box, lp is the distance from the tracking target to the outside of the perceived target box, ld is the distance from the tracking target to the inside of the virtual target box, and l l ,l w ,l h is the size of the virtual target box, δ is the state confidence level of the tracking target, and n is the frame difference between the perceived target and the tracking target; When the tracking target is outside the virtual target box, the confidence of the tracking target is The state credibility of the tracked target is δ = δ - n, where lo is the distance from the tracked target to the virtual target box, and lc is the distance from the tracked target to the perceived target box.

5. The method according to claim 1, wherein The determining the target minimum bounding box according to the position confidence and state confidence of the upsampled point includes: Calculating the first position confidence of the upsampled point in the perceived target point cloud cluster; Calculating the second position confidence of the upsampled point in the tracking target point cloud cluster; When the first position confidence and the second position confidence meet the same conditions, determining the target minimum bounding box.

6. The method according to claim 1, characterized in that, The 4D millimeter-wave radar is a radar-vision integrated machine.

7. A target tracking device, characterized in that, The device includes a processing module, a calculation module, and a target tracking module; The processing module is used for performing target perception using at least two 4D millimeter-wave radars. Each 4D millimeter-wave radar calculates target velocity information, clusters the collected point cloud to obtain a target point cloud cluster and the central position of the target point cloud cluster; The calculation module is configured to calculate the target heading angle and the target pitch angle based on the speed information of multiple frames of the same perceived target and the center position of the target point cloud cluster by using the stochastic parallel gradient descent algorithm; The target tracking module is configured to determine the target minimum bounding box according to the target heading angle and the target pitch angle obtained by the calculation module, so as to obtain the position information of the target; The target tracking module includes: An acquisition unit configured to acquire a tracking target list and a perceived target list, where the tracking target list includes target sampled point cloud and the probability prediction value corresponding to the point cloud, target Kalman state variables, Doppler speed information, and target ID, and the perceived target list includes target measured point cloud and Doppler speed information; An association unit configured to calculate the association between the tracking target and the perceived target according to the Hungarian algorithm to obtain the matching pairs of the tracking target and the perceived target; A first calculation unit configured to calculate the position of the upsampled point of the tracking target according to the upsampled historical position and the Doppler speed information; A second calculation unit configured to calculate the position credibility and the state credibility of the upsampled point according to the position of the upsampled point of the tracking target; A target tracking unit configured to determine the target minimum bounding box according to the position credibility and the state credibility of the upsampled point.

8. The device according to claim 7, wherein Calculating the position credibility and the state credibility of the upsampled point includes: Constructing a perceived target box and a virtual target box according to the tracking target list and the perceived target list; Determining the credibility and the state credibility of the tracking target according to the relationship between the tracking target and the perceived target box and the virtual target box.

9. A communication device, comprising a memory, a processor, and a program stored on the memory and executable on the processor; characterized in that, When the processor executes the program, it implements the target tracking method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the target tracking method according to any one of claims 1 to 6.

Images