Target tracking method, electronic device, and computer storage medium

Abstract

A target tracking method, an electronic device, and a computer storage medium, relating to the technical field of three-dimensional (3D) target tracking. The method can improve the 3D target tracking capability, and can be applied to a millimeter-wave radar. The millimeter-wave radar comprises a first radar and a second radar, and the first radar and the second radar have complementary detection orientations. The method comprises: determining a 3D predicted value corresponding to a target, a first predicted value, and a second predicted value; determining a corrected first gate and a corrected second gate; performing fusion matching on at least one first measured value falling within the first gate and at least one second measured value falling within the second gate to obtain a first 3D measured value; and obtaining a 3D trajectory on the basis of a plurality of first 3D measured values.

Description

Target tracking method, electronic device and computer storage medium

[0001] The present application claims priority to the Chinese patent application No. 202410949890.4, filed on July 15, 2024, and entitled "Target tracking method, electronic device and computer storage medium", the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] The present application relates to the technical field of millimeter wave radar, in particular to a target tracking method, an electronic device and a computer storage medium. BACKGROUND

[0003] Millimeter wave radar is a radar system that uses millimeter wave (usually refers to 30-300GHz frequency domain, corresponding to 1-10mm wavelength) electromagnetic waves for target detection and ranging.

[0004] At present, the tracking ability of millimeter wave radar for targets in three-dimensional space is weak, for example, for low, slow and small unmanned aerial vehicles and other targets, the ability of three-dimensional trajectory tracking is insufficient, resulting in low tracking accuracy and failing to meet the expected demand. SUMMARY

[0005] The present application discloses a target tracking method, an electronic device and a computer storage medium, which improves the tracking ability for targets in three-dimensional space by complementing the detection directions of multiple radars, and improves the accuracy of target tracking.

[0006] In a first aspect, the present application provides a target tracking method, which can be applied to a millimeter wave radar, the millimeter wave radar comprising a first radar and a second radar, the first radar and the second radar being complementary in detection direction; the method can determine a three-dimensional prediction value corresponding to a target; determine a first prediction value corresponding to the first radar and a second prediction value corresponding to the second radar; based on the three-dimensional prediction value and the first prediction value, obtain a corrected first gate; based on the three-dimensional prediction value and the second prediction value, obtain a corrected second gate; fuse and match at least one first measurement value falling into the first gate and at least one second measurement value falling into the second gate to obtain a first three-dimensional measurement value.

[0007] The first radar and the second radar are complementary in detection direction, specifically, the dimension of the two-dimensional measurement detected by the first radar relative to the three-dimensional measurement is different from the dimension of the two-dimensional measurement detected by the second radar relative to the three-dimensional measurement. The first radar and the second radar are considered to be complementary in detection direction if the measurement data obtained by the first radar and the second radar through the relative position layout design meet the above characteristics (i.e., the missing dimensions are different). For example, the beam plane formed by the antenna array of the first radar and the beam plane formed by the antenna array of the second radar are perpendicular or approximately perpendicular.

[0008] In some embodiments, the three-dimensional predicted value corresponding to the target can be obtained by a prediction algorithm such as a Kalman filter after obtaining the three-dimensional initial trajectory of the target.

[0009] In some embodiments, the first predicted value corresponding to the first radar and the second predicted value corresponding to the second radar can be obtained by a prediction algorithm such as a Kalman filter after obtaining the first two-dimensional initial trajectory corresponding to the first radar to obtain a two-dimensional first predicted value. The second two-dimensional initial trajectory corresponding to the second radar is obtained, and a prediction algorithm such as a Kalman filter is used to obtain a two-dimensional second predicted value.

[0010] The two-dimensional measurement is a short form of two-dimensional measurement data, and the three-dimensional measurement is a short form of three-dimensional measurement data. The two-dimensional measurement data includes two-dimensional measurement values of one point or multiple points. The three-dimensional measurement data includes three-dimensional measurement values of one point or multiple points. The measurement value or measurement data can refer to the data detected by the radar, and the predicted value or predicted data can be data obtained by predicting or calculating based on the measurement data. Unless otherwise specified, the two-dimensional or three-dimensional predicted value or measurement value refers to the two-dimensional or three-dimensional measurement value or predicted value corresponding to one point in the point cloud. For example, the two-dimensional measurement value can include two-dimensional coordinate values of one point. The three-dimensional measurement value can include three-dimensional coordinate values of one point in the point cloud. A 3D millimeter wave radar cannot directly detect three-dimensional measurement data, so the three-dimensional measurement value can be calculated by the method provided in the present application, rather than the measurement data directly detected by the radar.

[0011] It should be noted that, in order to prevent confusion, the three-dimensional measurement value obtained after the fusion matching of the first measurement value (two-dimensional) and the second measurement value (two-dimensional) falling into the corrected gate in the data association stage is defined as the first three-dimensional measurement value, and should not be understood as the 13th dimension measurement value in the order of 13. In the trajectory initiation stage (such as the track initiation stage), the three-dimensional measurement value obtained by trajectory matching (such as track matching) is defined as the second three-dimensional measurement value, and should not be understood as the 23rd dimension measurement value in the order of 23.

[0012] The first three-dimensional measurement value is a three-dimensional measurement value of a next point of the three-dimensional initial track. On the basis of the three-dimensional initial track, a three-dimensional measurement value of the next point (the first three-dimensional measurement value) is obtained in sequence, and the obtained next point is connected with the three-dimensional initial track, so that a track after the three-dimensional initial track is obtained.

[0013] Fusion matching refers to a process of performing matching calculation (for example, probability-based matching calculation or probability-based association calculation) on the two-dimensional first measurement value and the two-dimensional second measurement value falling into the modified gate, including a fusion process of generating a three-dimensional measurement (an initial value) from the two two-dimensional measurements, for example, three-dimensional coordinates obtained by converting two two-dimensional coordinates into the same coordinate system are three-dimensional measurement initial values. The fusion matching process involves data association and the fusion process of generating a three-dimensional measurement from two-dimensional measurements.

[0014] The method provided in the application can improve the accuracy of matching, obtain more accurate three-dimensional prediction values, and further improve the ability of 3D target tracking or the accuracy of tracking an entity target such as an unmanned aerial vehicle, by modifying the gate of the two-dimensional measurement according to the three-dimensional prediction value in the data association stage, and performing fusion matching between multiple points in the respective gates of different radars by using the modified gate. In addition, the fusion matching scheme only needs to perform matching between a limited number of points in the gate when calculating a three-dimensional measurement from two-dimensional measurements of multiple radars, thereby avoiding a series of problems caused by brute-force matching between point clouds.

[0015] In a possible implementation manner, before the three-dimensional prediction value corresponding to the tracking target is determined, matching of the first track and the second track can be further performed to determine that the first track and the second track correspond to the same target; the first track is one of at least one initial track corresponding to the first radar; the second track is one of at least one initial track corresponding to the second radar; and the three-dimensional initial track corresponding to the same target is obtained based on a plurality of first measurement values included in the first track and a plurality of second measurement values included in the second track.

[0016] In the track initiation stage, the measurement matching between points is converted into track matching between tracks, the response speed of track initiation is improved, and the initial track can be obtained more quickly. According to simulation tests, the conventional point-to-point matching method for determining the initial track in the related art generally causes a delay of minutes or tens of minutes, while the track matching method provided in the application can display the initial track in real time and synchronously, and the delay is only in the order of milliseconds or seconds. The track can be a motion track or a flight track, and the flight track can also be referred to as a track.

[0017] In a possible implementation, the matching of the first trajectory and the second trajectory is performed, and it is determined that the first trajectory and the second trajectory correspond to the same target can be that a first feature corresponding to the first trajectory and a second feature corresponding to the second trajectory are determined; and the first trajectory and the second trajectory correspond to the same target is determined according to a similarity between the first feature and the second feature.

[0018] In a possible implementation, the first feature and the second feature each include n sub-features; after the first feature corresponding to the first trajectory and the second feature corresponding to the second trajectory are determined, the method can further determine, based on the first feature and the second feature, feature coefficients respectively corresponding to the n sub-features; and the similarity between the first feature and the second feature is calculated based on the feature coefficients.

[0019] In a possible implementation, the first measurement value detected by the first radar includes a two-dimensional coordinate that lacks a third dimension coordinate; the second measurement value detected by the second radar includes a two-dimensional coordinate that lacks a first dimension or a second dimension; and the three-dimensional initial trajectory corresponding to the same target is obtained based on the first measurement value included in the first trajectory and the second measurement value included in the second trajectory, including: obtaining the third dimension coordinate that is missing in the first measurement value based on the two-dimensional coordinate included in the second measurement value and the two-dimensional coordinate included in the first measurement value; and combining the two-dimensional coordinate included in the first measurement value and the obtained third dimension coordinate to obtain a second three-dimensional measurement value; and the three-dimensional initial trajectory is obtained based on the plurality of second three-dimensional measurement values corresponding to the same target.

[0020] As described previously, the two-dimensional coordinate in the first measurement value lacks a dimension compared to a three-dimensional space coordinate, and the two-dimensional coordinate in the second measurement value lacks a dimension compared to a three-dimensional space coordinate are different, for example, the first measurement value includes a first dimension coordinate and a second dimension coordinate, and lacks a third dimension; the second measurement value includes a first dimension coordinate and a third dimension coordinate, and lacks a second dimension coordinate; or the second measurement value includes a second dimension coordinate and a third dimension coordinate, and lacks a first dimension coordinate. The two-dimensional coordinate in the first measurement value is a coordinate recorded based on a first coordinate system corresponding to the first radar; and the two-dimensional coordinate in the second measurement value is a coordinate recorded based on a second coordinate system corresponding to the second radar.

[0021] Specifically, based on the two-dimensional coordinates included in the second measurement value and the two-dimensional coordinates included in the first measurement value, the third dimension coordinate missing in the first measurement value can be obtained according to the first dimension coordinate and the second dimension coordinate included in the first measurement value, and according to the first dimension coordinate and the third dimension coordinate included in the second measurement value or according to the second dimension coordinate and the third dimension coordinate included in the second measurement value, by conversion, the third dimension coordinate missing in the first coordinate system corresponding to the first radar is obtained. In this way, the two-dimensional coordinates of the existing two dimensions in the first measurement value are combined with the obtained third dimension coordinates, that is, three-dimensional measurement values (second three-dimensional measurement values) of three dimensions are obtained. For example, the two-dimensional coordinates included in the first measurement value are (r, θ, *), and the two-dimensional coordinates included in the second measurement value are (r, *, φ), or the two-dimensional coordinates in the first measurement value are (*, θ, φ), and the two-dimensional coordinates in the second measurement value are (r, *, φ) or (r, θ, *), as long as the missing dimensions of the two two-dimensional coordinates are different. A three-dimensional measurement value is a measurement value of a point, so that three-dimensional measurement values of multiple points can be obtained, and adjacent points in the multiple points are connected by lines, so that a trajectory is obtained, and the obtained trajectory is a three-dimensional starting trajectory.

[0022] Among them, in time sequence, the second three-dimensional measurement value is obtained before the first three-dimensional measurement value, the second three-dimensional measurement value is used to determine the three-dimensional starting trajectory, and the first three-dimensional measurement value is used to determine the trajectory after the three-dimensional starting trajectory.

[0023] In a possible implementation, after obtaining the three-dimensional starting trajectory corresponding to the same target, the acceleration values corresponding to the first dimension, the second dimension and the third dimension can be calculated according to the first measurement value and the second measurement value; the three dimensions corresponding to the speed and the position of the three-dimensional starting trajectory are respectively corrected by using the acceleration values, to obtain the corrected speed information and the position information.

[0024] According to the first measurement value and the second measurement value, the acceleration values corresponding to the first dimension, the second dimension and the third dimension can be obtained by averaging the first acceleration corresponding to the first radar and the second acceleration corresponding to the second radar, that is, according to the first measurement value and the second measurement value, the acceleration average (average of acceleration components) of the first radar and the second radar in three dimensions is calculated. The first acceleration can be obtained according to the speed information in the first measurement value, for example, the acceleration is obtained according to the speed information in the first measurement value of two adjacent points (for example, two points of adjacent frames), as the first acceleration corresponding to one of the points. Similarly, the second acceleration can be obtained according to the speed information in the second measurement value. In other embodiments, the acceleration values corresponding to the first dimension, the second dimension and the third dimension can be obtained by weighted sum of the first acceleration and the second acceleration, for example, the weight corresponding to the first radar is 0.48, the weight corresponding to the second radar is 0.52, and the like.

[0025] The acceleration values obtained by integrating multiple radars are used to correct the three-dimensional starting trajectory, which can correct the influence of large single-radar measurement error on the accuracy of the three-dimensional starting trajectory to a certain extent, and improve the accuracy of the obtained three-dimensional starting trajectory.

[0026] In a possible implementation, the three-dimensional prediction value corresponding to the tracking target can be obtained by performing prediction based on the three-dimensional starting trajectory.

[0027] In a possible implementation, the corrected first gate can be obtained based on the three-dimensional prediction value and the first prediction value, that is, the corrected first prediction value is obtained based on the three-dimensional prediction value and the two-dimensional first prediction value corresponding to the same point, and the corrected first gate is obtained according to the corrected first prediction value. The first gate is a gate centered on the corrected first prediction value.

[0028] In a possible implementation, the corrected first prediction value can be obtained based on the three-dimensional prediction value and the two-dimensional first prediction value corresponding to the same point, that is, the corrected first prediction value is obtained according to the three-dimensional prediction value and the first prediction value and a weighting coefficient. The weighting coefficient is used to control the proportion of the three-dimensional prediction value or the first prediction value.

[0029] In a possible implementation, the fusion matching of the at least one first measurement value falling into the first wave gate and the at least one second measurement value falling into the second wave gate can be: determining m points falling into the first wave gate; determining n points falling into the second wave gate; and performing fusion matching on the m points and the n points based on probabilities corresponding to the m points and the n points, to obtain a first three-dimensional measurement value.

[0030] In a possible implementation, after obtaining the associated three-dimensional trajectory, the two-dimensional first measurement value corresponding to the first radar can be determined in a case where the first radar does not lose frames and the second radar loses frames, and the first three-dimensional measurement value is corrected according to the first measurement value.

[0031] In a second aspect, an embodiment of the present application further provides a target tracking method, which can be applied to a millimeter wave radar including a first radar and a second radar, and the first radar and the second radar are complementary in a detection direction. The method includes: performing matching of a first trajectory and a second trajectory to determine that the first trajectory and the second trajectory correspond to a same target; the first trajectory is one of at least one starting trajectory corresponding to the first radar; the second trajectory is one of at least one starting trajectory corresponding to the second radar; and a three-dimensional starting trajectory corresponding to the same target is obtained based on a plurality of first measurement values included in the first trajectory and a plurality of second measurement values included in the second trajectory.

[0032] Converting the matching between points corresponding to different radars into the matching between trajectories corresponding to different radars can accelerate the starting speed of the track, shorten the starting speed of the track from minutes (for example, several minutes or tens of minutes) to milliseconds or seconds, and improve the accuracy and stability of the starting of the three-dimensional track.

[0033] In a third aspect, the present application further provides a target tracking method, which can be applied to a millimeter wave radar including a first radar and a second radar, and the first radar and the second radar are complementary in a detection direction. The method includes: determining a two-dimensional first measurement value corresponding to the first radar in a case where the first radar does not lose frames and the second radar loses frames; and correcting a first three-dimensional measurement value according to the first measurement value; the first three-dimensional measurement value is obtained based on a first measurement value corresponding to the first radar and a second measurement value corresponding to the second radar.

[0034] In the track completion phase, the two-dimensional measurement value not losing frames is used to correct the three-dimensional measurement value (not actually measured, but calculated, which can also be referred to as a three-dimensional prediction value), so that a more accurate three-dimensional measurement value can be obtained, and a large error caused by continuous frame loss can be avoided.

[0035] In a fourth aspect, the present application provides an electronic device, comprising a memory for storing a computer program; and a processor for executing the computer program in the memory to perform the target tracking method provided in the first aspect and any one of the implementation manners of the first aspect. For example, the electronic device can be a millimeter wave radar.

[0036] In a fifth aspect, the present application provides a computer storage medium, which stores a computer program, and the computer program is executed by a processor to implement the target tracking method provided in the first aspect and any one of the implementation manners of the first aspect.

[0037] In a sixth aspect, the present application provides a computer program product, which, when running on an electronic device, causes the electronic device to perform the target tracking method provided in the first aspect and any one of the implementation manners of the first aspect.

[0038] In a seventh aspect, the present application provides an electronic device, which comprises the method or device introduced in any one of the aspects or implementation manners of the present application. The electronic device can be a chip.

[0039] It should be understood that the description of technical features, technical solutions, advantages or similar language in the present application does not imply that all features and advantages can be realized in any single embodiment. On the contrary, it can be understood that the description of a feature or advantage means that the specific technical feature, technical solution or advantage is included in at least one embodiment. Therefore, the description of technical features, technical solutions or advantages in the present application does not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions and advantages described in the present application can be combined in any appropriate manner. Those skilled in the art will understand that the present application can be implemented without one or more specific technical features, technical solutions or advantages of a particular embodiment. In other embodiments, additional technical features and advantages can be identified in specific embodiments that do not embody all embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0040] The following describes the drawings used in the present application.

[0041] FIG. 1 is a schematic diagram of a two-dimensional plane divided into regions in the related art;

[0042] FIG. 2 is a schematic diagram of the hardware architecture of a millimeter wave radar;

[0043] FIG. 3 is a schematic diagram of two radars detecting complementary directions in some embodiments of the present application;

[0044] FIG. 4 is an example of an application scenario of the present application;

[0045] FIG. 5 is an example diagram of a system architecture corresponding to some embodiments of the present application;

[0046] FIG. 6 is a flow processing architecture diagram corresponding to some embodiments of the present application;

[0047] FIG. 7 is a diagram of brute force matching;

[0048] FIG. 8 is a diagram of a track initiation stage in some embodiments of a target tracking method provided by the present application;

[0049] FIG. 9 is a diagram of a three-dimensional prediction feedback to two-dimensional measurement data cycle mechanism in some embodiments of a target tracking method provided by the present application;

[0050] FIG. 10 is a diagram of a data association stage in some embodiments of a target tracking method provided by the present application;

[0051] FIG. 11 is a diagram of obtaining a predicted value from a measurement value in some embodiments of a target tracking method provided by the present application;

[0052] FIG. 12 is a diagram of gate correction in some embodiments of a target tracking method provided by the present application;

[0053] FIG. 13 is a diagram of determining a measurement point according to a corrected gate in some embodiments of a target tracking method provided by the present application;

[0054] FIG. 14 is a diagram of fusion matching of m points in a gate and n points in another gate in some embodiments of a target tracking method provided by the present application. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present application will be described below with reference to the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0056] In the description of the embodiments of the present application, unless otherwise specified, " / " represents the meaning of or, for example, A / B can represent A or B. The "and / or" in the text only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, in the description of the embodiments of the present application, "multiple" means two or more than two.

[0057] Hereinafter, the terms "first", "second", etc. are used only for the purpose of description, and should not be construed as implying or suggesting relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features.

[0058] In the embodiments of the present application, the words "exemplary" or "for example" are used to mean serving as an example, instance, or illustration, in the embodiments of the present application. Any embodiment or design scheme described in the embodiments of the present application as "exemplary" or "for example" should not be construed as being more preferred or advantageous than other embodiments or design schemes.

[0059] Millimeter wave radar is a radar working in the millimeter wave band (millimeter wave). Generally, millimeter wave refers to a frequency domain of 30-300 GHz (wavelength of 1-10 mm), and the wavelength of millimeter wave is between that of microwave and centimeter wave. Millimeter wave radar can be applied to 3D (Three-dimensional) target tracking, for example, can be used to track unmanned aerial vehicles and other physical targets.

[0060] For example, taking unmanned aerial vehicles as tracking targets, according to statistics, in a specified area, there can be a large number of non-cooperative or abnormal unmanned aerial vehicles, which need to be tracked for various reasons. Through millimeter wave radar, it is expected to achieve accurate and stable tracking of unmanned aerial vehicles under full-time, all-weather, and long-distance conditions, realize unmanned aerial vehicle perception, and provide an effective auxiliary tool for the management of specified spaces or areas (such as specified areas in a city, an airport), etc.

[0061] However, the current scheme for 3D tracking of physical targets such as unmanned aerial vehicles based on millimeter wave radar has limited tracking capability and is difficult to meet the actual management needs.

[0062] For example, related technical scheme one proposes a multi-radar track association method based on region division, as shown in FIG. 1. The idea of region division is used to divide multiple regions in a two-dimensional plane, which are numbered respectively, for example, multiple regions numbered from 00010003 to 10031003 are shown in FIG. 1. In this way, multiple target tracks can be divided in different regions, and subsequent association calculations can be based on track region numbers to avoid interference from unrelated tracks.

[0063] This scheme one only performs track association on the target in the current two-dimensional plane, and can only process two-dimensional plane data. In actual application scenarios, the target to be tracked is usually a three-dimensional entity target, and the track is three-dimensional data. This scheme lacks tracking capability for 3D targets, lacks processing of missing dimension measurement data, and is difficult to meet the requirements of track tracking technology for unmanned aerial vehicles and other entity targets. Missing dimension measurement refers to measurement data missing one dimension in three dimensions of space. For example, the measurement data detected by a single 3D millimeter wave radar is two-dimensional data, which is missing one dimension compared with three-dimensional data, and thus is missing dimension measurement.

[0064] For another example, the related technical scheme two proposes a 3D target tracking method based on a two-radar two-coordinate external source radar system. The target height interval is divided into multiple height subintervals, then the weight of the target in each height subinterval is calculated by using the measurement value recursion, and finally the target height is estimated by weighted fusion of each height subinterval.

[0065] This scheme two only considers the track tracking of a target with constant height (fixed height interval), i.e., the target only moves in a fixed horizontal plane. However, in actual applications, the target to be tracked is usually free to fly and does not maintain at a fixed height in the horizontal plane. Therefore, this scheme two cannot track a target free to fly in a 3D plane, and does not involve the association problem of multiple targets in a 3D space.

[0066] As can be seen, in the related art, the scheme for tracking a 3D target based on a millimeter wave radar has insufficient tracking capability. One of the reasons for the insufficient tracking capability is that the antenna distribution of a single radar is one-dimensional linear array, which can only obtain two-dimensional information, and the measurement space is two-dimensional, while the target state space is three-dimensional. The dimensions of the measurement space and the target state space do not match. The insufficient tracking capability can be embodied in at least one of the following three aspects:

[0067] (1) Difficulty in track initiation: The missing dimensions of two-dimensional measurement of radars at different detection directions are inconsistent, and it is difficult to match in a three-dimensional space, so it is difficult to initiate a track.

[0068] (2) Low accuracy of data association after track initiation: A single radar lacks one dimension, and the predicted point based on the missing dimension measurement has a large error, so the accuracy of data association is low.

[0069] (3) Frame loss of measurement data detected by a single radar: When tracking a small-volume entity target at a long distance in a three-dimensional space, greater jitter may occur, frame loss phenomenon is easy to occur, the possibility of measurement loss (frame loss) is higher, and continuous tracking is more difficult.

[0070] In view of this, the embodiment of the present application proposes a tracking method, at least two radars are deployed, and information complementation is formed in azimuth and pitch by the two radars, and multi-radar cooperation is used to complete target three-dimensional trajectory fusion tracking. In response to the technical problem of insufficient tracking capability, at least one of the following three technical means is proposed: three-dimensional track initiation technology based on two-dimensional track initial matching, gate matching association technology based on three-dimensional trajectory feedback, and three-dimensional trajectory completion technology based on two-dimensional measurement guidance.

[0071] In some embodiments, the three-dimensional track initiation technology based on two-dimensional track initial matching can be to use the two-dimensional tracks of multiple single radars for initial matching, filter out single radar clutter, and use track initial matching to achieve accurate matching of two-dimensional measurement points. That is, using the matched two-dimensional measurement, three-dimensional measurement is generated, and three-dimensional track initiation is completed.

[0072] In some embodiments, the gate matching association technology based on three-dimensional trajectory feedback can be to predict the predicted value of the three-dimensional trajectory according to the three-dimensional track initiation, feed back the predicted value to the single radar, update the associated gate, use the updated gate to accurately associate the two-dimensional measurement, and obtain the measurement value of the three-dimensional trajectory through the three-dimensional probability weighted tracking algorithm, thereby improving the data association accuracy and further improving the three-dimensional tracking accuracy.

[0073] In some embodiments, the three-dimensional trajectory completion technology based on two-dimensional measurement guidance can be to correct and complete the global three-dimensional trajectory by using the observed two-dimensional measurement of the single radar in response to the frame loss phenomenon of the single radar.

[0074] The target tracking method proposed in the embodiment of the present application can be applied to 3D target tracking or automatic driving of a car and other application scenarios. For example, it can also be applied to tracking scenarios of vehicle, ship and other entity targets, and can be deployed in designated places such as airports to detect surrounding interference. The following will mainly take the 3D target tracking scenario as an example for description. For example, FIG. 4 shows an example of a typical 3D target tracking scenario: a UAV tracking scenario.

[0075] The target tracking method proposed in the embodiment of the present application can obtain various specific forms of products. For example, the product form can be a computer program product carried on a physical storage medium or transmitted through a network, or can be a millimeter wave radar loaded with a corresponding computer program. For example, the specific product can be a single radar or a radar and vision integrated machine, a millimeter wave radar system, or part of the components of the whole machine.

[0076] The target tracking method proposed in the embodiment of the present application can be implemented based on a millimeter wave radar and other devices. The millimeter wave radar can be deployed in a building or a designated control area to track entity targets such as UAVs, and realize 3D position tracking and reporting of UAVs and other targets.

[0077] Exemplarily, the method proposed in the embodiments of the present application can be applied to 3D millimeter wave radars in some embodiments, and can also be applied to 4D millimeter wave radars with low elevation resolution in other embodiments.

[0078] Exemplarily, the system architecture of a single 3D millimeter wave radar is shown in FIG. 2. The 3D millimeter wave radar includes an antenna array, a transmitter, a voltage controlled oscillator (VCO), a power amplifier (PA), an operation unit, and an electronic control unit (ECU).

[0079] In the embodiments of the present application, at least two radars form information complementarity in azimuth and elevation. The antenna arrays of the two radars in the at least two radars can be perpendicular to each other, or in other words, the beam directions formed by the antenna arrays of the two radars are perpendicular to each other. Exemplarily, as shown in FIG. 3, the at least two radars include a first radar and a second radar. In the structural design, the antenna array of the first radar is vertically distributed relative to the antenna array of the second radar, for example, the antenna array of the first radar forms a beam in the elevation direction, and the antenna array of the second radar forms a beam in the azimuth direction.

[0080] Exemplarily, in other embodiments, a third radar or a fourth radar can also be deployed. The antenna array of the third radar forms a beam direction perpendicular to the beam direction formed by the antenna array of the first radar or the second radar. The fourth radar forms a beam direction perpendicular to the beam direction formed by the antenna array of at least one of the first radar to the third radar.

[0081] The at least two radars forming information complementarity in azimuth and elevation can be deployed in the same shell or can be independently set respectively. The deployment in the same shell means that the hardware of the radar is internally composed of multiple independent chip arrays, or in other words, at least two groups of 3D millimeter wave radar circuits are built in the same shell. The main structure of one independent chip array or one group of 3D millimeter wave radar circuits can be as shown in FIG. 2, and the at least two groups of 3D millimeter wave radar circuits can share the same ECU. The independent setting respectively means that the at least two radars are spaced apart by a certain distance and are deployed in different detection positions. The beam formed by the antenna arrays of the two radars is perpendicular to each other.

[0082] The beam direction formed by the antenna array of a single 3D millimeter wave radar is a plane, and the measurement data obtained by the point cloud detection is two-dimensional data, that is, only two dimensions of three-dimensional data are included. The point cloud is composed of a plurality of reflection points, and the measurement value of each reflection point includes two-dimensional coordinates and velocity information. The velocity is the radial velocity of the target relative to the radar.

[0083] The two-dimensional coordinates can be represented based on a Cartesian coordinate system or based on a polar coordinate system. For example, in the rectangular coordinate system, the two-dimensional coordinates detected by a single radar are (x, y) or (x, z) or (y, z). In the polar coordinate system, the two-dimensional coordinates detected by a single radar are (r, θ) or (r, φ). The distance r represents the range, the azimuth θ represents the azimuth, and the elevation φ represents the elevation. The distance is the straight-line distance between the target object and the radar. The azimuth and the elevation respectively represent the azimuth angle or the elevation angle of the target object relative to the radar in the horizontal plane or the vertical plane.

[0084] In the embodiments of the present application, at least two radars are deployed, and the two radars are complementary in the detection direction. Specifically, the two-dimensional coordinates detected by the first radar include the first dimension and the second dimension in three dimensions, and the two-dimensional coordinates detected by the second radar include at least the third dimension in three dimensions, and further include one of the first dimension and the second dimension. For example, taking the rectangular coordinate as an example, in the case that the two-dimensional coordinates of the measurement value detected by the first radar are (x, y), the two-dimensional coordinates of the measurement value detected by the second radar are (x, z) or (y, z). Taking the polar coordinate as an example, the two-dimensional coordinates detected by the first radar are (r, θ), and the two-dimensional coordinates detected by the second radar are (r, φ).

[0085] In combination with the actual application scenario, one of the typical application scenarios of the target tracking method proposed in the embodiments of the present application is a 3D target tracking scene. As shown in FIG. 4, taking a drone as an example, the millimeter wave radar includes a first radar and a second radar, and the first radar and the second radar are complementary in the detection direction. Specifically, the second beam plane formed by the antenna array of the second radar is perpendicular to the first beam plane formed by the antenna array of the first radar, so that the measurement data detected by the two radars can be complementary. For example, as shown in FIG. 4, taking the polar coordinate as an example, the measurement data detected by the first radar includes two-dimensional coordinates (r, θ) and velocity V, and the measurement data detected by the second radar includes two-dimensional coordinates (r, φ) and velocity V. The two-dimensional coordinates are used to represent the position of the target object, that is, the position of the drone.

[0086] A specific embodiment is listed below. The embodiment takes an example of a millimeter wave radar system (or simply millimeter wave radar) including two radars, and an example of a detection target being a UAV to illustrate. In other embodiments, more radars can be provided, and dimension reduction matching is performed according to the method described in the embodiments of the present application.

[0087] Exemplarily, the target tracking method proposed in the embodiments of the present application can be executed based on the system architecture shown in FIG. 5. The millimeter wave radar includes a horizontal radar and a vertical radar. The horizontal radar, that is, the antenna array of the radar is horizontal, or the beam emitted by the antenna array is parallel to the horizontal plane; the vertical radar, that is, the antenna array of the radar is vertical, or the beam emitted by the antenna array is perpendicular to the horizontal plane.

[0088] It should be noted that the horizontal radar and the vertical radar are only an example of complementary detection directions between the two radars, and in actual application, the beam planes of the two radars can be perpendicular to each other in the three-dimensional space, and are not limited to being parallel or perpendicular to the horizontal plane.

[0089] As shown in FIG. 5, the data processing of 3D target tracking can include four stages of plot condensation, track initiation, data association, and track generation. In the track initiation stage, the embodiments of the present application propose a three-dimensional track initiation technology based on two-dimensional track preliminary matching, which can specifically include three sub-stages of constructing a two-dimensional track feature, track preliminary matching, and measurement accurate matching.

[0090] In the data association stage, the embodiments of the present application propose a gate matching association technology based on three-dimensional track feedback, which can specifically include two sub-stages of 2D gate correction and 3D target tracking after gate matching.

[0091] In the track completion stage, the embodiments of the present application propose a three-dimensional track completion technology based on two-dimensional measurement guidance, which can specifically include three-dimensional track completion.

[0092] As shown in FIG. 6, exemplarily, it is assumed that the two-dimensional coordinate information of each reflection point in the point cloud data collected by the horizontal radar is T(x, z), and the two-dimensional coordinate information of each reflection point in the point cloud data collected by the vertical radar is T(x, y, z). The target tracking method proposed in the embodiments of the present application realizes three-dimensional track initiation by fast matching of the two dimensionally reduced point clouds, or dimensionally reduced matching of the point cloud data detected by the two radars, performs data association based on the initiated track, and completes the three-dimensional track after the three-dimensional track is obtained by data association, to compensate for the frame loss phenomenon of the radar.

[0093] It should be noted that the track refers to the track of the target object driving or flying or moving, and the track can include a flight path, for example, in the example of a UAV as a target, the track is a flight path.

[0094] It should be noted that in the related art, the matching of multiple radar data is only in a two-dimensional plane, and the coordinates in the point cloud data detected by multiple radars are converted into a two-dimensional coordinate system for matching. The target in the 3D space has three-dimensional position information, and the input information of two radars has only two dimensions, which cannot directly locate the target in the 3D space according to the two-dimensional information of the two radars, and it is necessary to convert the two-dimensional point cloud detected by the radar into a three-dimensional coordinate system. If a brute-force matching is directly performed on the two-dimensional point cloud data of the two radars, the cost is too large and it is difficult to implement. The method proposed in the embodiments of the present application can realize fast matching of the missing dimension point cloud of the two radars.

[0095] The target tracking method of the present embodiment will be described below in three stages of track initiation, data association and track generation.

[0096] Track initiation stage:

[0097] The two radars each lack different dimensions and cannot be unified in the same three-dimensional coordinate system, or in other words, cannot be matched in the same coordinate system. The brute-force matching method, that is, directly generating a full permutation of the missing dimension point tracks of the two radars, as shown in FIG. 7, assumes that the point cloud data (i.e., the R1 point track shown in FIG. 7) detected by the first radar R1 includes measurement data of M transmitting points, and the point cloud data (the R2 point track shown in FIG. 7) detected by the second radar R2 includes measurement data of N transmitting points. The brute-force matching will form M*N matching hypotheses, and then use distance and azimuth information for screening. This matching method has a large number of point clouds, and it is difficult to match and screen, and cannot be tracked in real time. For example, in high-density point cloud imaging, the point cloud detected by a single frame of radar data can include 3-4 ten thousand reflection points. The brute-force matching of a single frame of data can have 30000*30000 matching hypotheses, and multiple frames of data need to face more matching hypotheses, which consumes a lot of computing resources and is difficult to implement.

[0098] The embodiments of the present application propose to convert the matching of points into the matching of tracks. For example, the above brute-force matching method is the matching between points and points, specifically the matching between the measurement data of a point detected by one radar and the measurement data of a point detected by another radar.

[0099] The embodiments of the present application propose to use the two-dimensional tracks of multiple single radars for initial matching, filter out single radar clutter, and use track matching to realize accurate matching of two-dimensional measurement points and three-dimensional track initiation.

[0100] Specifically, in the present embodiment, as shown in FIG. 8, the matching of tracks can include the following steps:

[0101] S101, constructing track features:

[0102] The track features can include length, position change R gap , and speed difference V gapat least one of a length feature, a position change feature, a velocity difference feature, a history acceleration feature, a trajectory curvature, historical acceleration, trajectory curvature, and the like.

[0103] The length feature is a length of a flight path formed by the target object, such as a UAV, during flight, which is generally a length of an irregular curve or a regular curve between an end point and a start point.

[0104] Exemplarily, the length feature of the flight path can be determined according to point cloud data of a last frame (up to the current time) detected by the radar and point cloud data of a first frame. For example, a point is determined from the point cloud of the last frame as an end point, and a point is determined from the point cloud of the first frame as a start point, and a length between the end point and the start point is the length of the flight path. The length between the end point and the start point can be obtained by integral operation, that is, a difference between distances of adjacent two points is accumulated and summed to obtain a length value. The adjacent two points are two points determined from adjacent two frames of point cloud respectively.

[0105] The point determined from a frame of point cloud can be determined in various ways, for example, a point is randomly selected, or a mean value of measurement data of each point in the current frame is calculated. For example, when calculating the length, a mean value of two-dimensional coordinates of each point in a frame of point cloud is calculated, and the calculated mean value is used as position information of the determined point. The length of the flight path is calculated based on the points determined from each frame of point cloud respectively.

[0106] The position change feature represents a position difference between the end point and the start point up to the current time, which can be a difference between two-dimensional coordinates of the end point and two-dimensional coordinates of the start point. The determination method of the start point and the end point can refer to the method in the length feature, which is not described herein.

[0107] The velocity difference represents a velocity difference between the end point and the start point up to the current time, which can be a difference between a velocity of the end point and a velocity of the start point.

[0108] The history acceleration represents a mean value of accelerations between adjacent two points in a point cloud corresponding to a target object detected at the current time. For example, adjacent two points in M reflection points are a group, and an acceleration value can be determined. M points correspond to M-1 groups, and M-1 accelerations are averaged to obtain the history acceleration. The adjacent two points are a group, and the acceleration value can be calculated as a difference between velocities of adjacent two points in adjacent two frames of point cloud, and the acceleration value corresponding to the adjacent two points is calculated according to a time interval between the adjacent two frames and the difference between the velocities of the adjacent two points.

[0109] The trajectory curvature represents the reciprocal of the radius of the fitted curve. In this embodiment, the initial fitted curve can be obtained by preliminary fitting based on multiple frames of point cloud data. The fitted curve is segmented, and the curvature of each segment is calculated. The average curvature or the sum of curvatures is taken as the curvature of the entire trajectory.

[0110] Specifically, in this embodiment, the target object is a UAV, denoted as target1, the first radar is denoted as tr1, and the second radar is denoted as tr2. The track features are constructed based on the length, position change, speed difference, historical acceleration, and trajectory curvature of the UAV track.

[0111] wherein, denotes the track feature corresponding to the first radar, denotes the first feature to the Nth feature constructed based on the point cloud of the UAV, for example, denotes the length feature, denotes the trajectory curvature.

[0112] Similarly, the track feature corresponding to the second radar can be constructed as

[0113] S102, performing matching between the track corresponding to the first radar and the track corresponding to the second radar.

[0114] In actual application scenarios, the first radar can detect at least one target, and at least one track feature can be constructed accordingly. Similarly, the second radar can detect at least one target, and at least one track feature can be constructed accordingly.

[0115] After obtaining at least one track feature corresponding to the two radars respectively, the similarity between the track features of different radars can be calculated, and the tracks of different radars can be matched according to the similarity.

[0116] In this embodiment, a feature weighting based track matching is proposed, that is, feature weighting is used in the similarity calculation adopted in this embodiment.

[0117] The features obtained in S101 are subjected to dimensionless processing, and the weighting coefficients p of each feature are calculated.

[0118] wherein, p n denotes the nth feature in the N features, denotes the nth feature corresponding to the first radar, denotes the nth feature corresponding to the second radar, where i represents the i-th feature, and the i-th feature is one of the N features other than the n-th feature, represents the i-th feature corresponding to the first radar, represents the i-th feature corresponding to the second radar, min represents the minimum value, max represents the maximum value. μ is a hyperparameter, for example, μ ∈ [0, 1].

[0119] In this way, the feature weighting coefficient corresponding to each of the N features can be obtained.

[0120] Next, the feature weighting coefficient is used to calculate the track similarity δ of each track of the two radars, and track matching is performed. If both radars detect only one target, the similarity calculation formula is as follows:

[0121] wherein, represents the similarity between the track features corresponding to the first radar and the track features corresponding to the second radar, and respectively represent the 1st feature and the n-th feature corresponding to the only target detected by the first radar, and respectively represent the 1st feature and the n-th feature corresponding to the only target detected by the second radar.

[0122] If the two radars detect multiple targets respectively, the similarity calculation formula is as follows:

[0123] represents the p-th track feature corresponding to the first radar, represents the q-th track feature corresponding to the second radar, respectively represent the 1st feature and the n-th feature in the p-th track feature corresponding to the first radar. respectively represent the 1st feature and the n-th feature in the q-th track feature corresponding to the second radar.

[0124] It should be noted that the way of calculating the similarity between track features is not limited to the above calculation method based on feature weighting. For example, in other embodiments, one or more of the following similarity algorithms can be used: Euclidean distance, Manhattan distance, cosine similarity, Mahalanobis distance, etc.

[0125] Thus, after obtaining the at least one track feature corresponding to the first radar and the at least one track feature corresponding to the second radar, the similarity between each two track features can be calculated, and the track pairs with high similarity can be screened out according to the similarity, and the track pairs with high similarity are considered to be matched. The similarity is high, that is, the similarity exceeds a predetermined threshold, for example, the similarity exceeds 0.9 or 90%, and the track pairs are considered to be matched. The two tracks from different radars are matched, which means that the two tracks come from the same target or are tracks of the same target.

[0126] In S103, the matched track pairs are used to perform matching of two-dimensional measurements to generate three-dimensional measurements.

[0127] For example, assuming that the pth track corresponding to the first radar is matched with the qth track matched by the second radar, then the association of the two-dimensional measurement data (abbreviated as two-dimensional measurement) corresponding to the pth track and the two-dimensional measurement data corresponding to the qth track is performed to generate three-dimensional measurement data.

[0128] Specifically, the matching between the two-dimensional measurement of the pth track and the two-dimensional measurement of the qth track includes the matching between the two-dimensional coordinates of the two radars. For example, the specific matching manner can be to project the two-dimensional coordinates of one radar into the coordinates of the other radar to generate three-dimensional coordinates.

[0129] For example, the two-dimensional measurement data corresponding to the second radar includes two-dimensional coordinates and velocity information, for example, the two-dimensional coordinates are (r, θ). The measurement data corresponding to the first radar includes two-dimensional coordinates and velocity information, for example, the two-dimensional coordinates are (r, φ). In this embodiment, for the convenience of calculation, the two-dimensional measurement data is expressed in three-dimensional form, for example, the two-dimensional coordinates of the measurement data corresponding to the first radar are (r, *, φ), and the two-dimensional coordinates of the measurement data corresponding to the second radar are (r, θ, *), where * represents the missing dimension.

[0130] The two-dimensional coordinates of the first radar are projected into the coordinate system of the second radar to obtain the value of the missing dimension of the second radar through conversion; or the two-dimensional coordinates of the second radar are projected into the coordinate system of the first radar to obtain the value of the missing dimension of the first radar through conversion. For example, the missing dimension of the second radar is the elevation angle φ, and the two-dimensional coordinates (r, φ) of the first radar are projected into the coordinate system of the second radar to calculate the missing φ of the second radar according to (r, φ) to obtain the three-dimensional coordinates (r, θ, φ) of the second radar. Thus, the obtained three-dimensional coordinates and velocity information constitute the three-dimensional measurement data (hereinafter referred to as three-dimensional measurement).

[0131] It should be noted that the matching of the two-dimensional coordinates is the matching calculation between two points of the same timestamp.

[0132] In this embodiment, after the three-dimensional measurement is generated, the generated three-dimensional measurement is also corrected, and a three-dimensional initial track (or three-dimensional initial trajectory) is obtained according to the corrected measurement data. In other embodiments, the three-dimensional measurement can not be corrected, but a three-dimensional initial track is directly obtained based on the generated three-dimensional measurement data.

[0133] S104, correcting the three-dimensional measurement to obtain a three-dimensional initial track.

[0134] The position (i.e., two-dimensional coordinates) and velocity information in the two-dimensional measurement data (or simply two-dimensional measurement) corresponding to the two radars are used to calculate the average acceleration in each of the three dimensions

[0135] Where t represents the dimension, m represents the number of radars, represents the average acceleration in the tth dimension, n represents the nth point, n-1 represents the n-1th point, and h and s represent different radars, for example, h represents the first radar and s represents the second radar. represents the component of the velocity vector of the nth point corresponding to radar h in the tth dimension. T represents the time interval between two adjacent frames, for example, the interval between receiving two adjacent echoes. represents the component of the velocity vector of the nth point corresponding to radar h in the tth dimension.

[0136] For example, t = r, then

[0137] represents the component of the velocity vector of the nth point corresponding to radar h in the rth dimension.

[0138] Next, the average acceleration in each dimension is used to correct the target velocity and the three-dimensional position

[0139] Using the average acceleration in each dimension, the target velocity and three-dimensional position in each dimension are corrected to obtain an accurate three-dimensional target position:

[0140] Where V t represents the component of the velocity vector of the three-dimensional measurement data before updating in the tth dimension, represents the component of the velocity vector of the three-dimensional measurement data after updating in the tth dimension, P t ​a value of a three-dimensional coordinate representing the three-dimensional measurement data before updating in a t-th dimension, a value of a three-dimensional coordinate representing the three-dimensional measurement data after updating in a t-th dimension. K is a hyperparameter, for example, k ∈ [0, 1]. g a is a function for mapping acceleration to velocity, p a is a function for mapping acceleration to position coordinates, g a and p a There can be multiple implementations, exemplarily,

[0141] T represents a time interval between two adjacent echoes received.

[0142] Thus, the updated velocity components and coordinate values in the r dimension, the θ dimension and the φ dimension respectively can be obtained, and according to the velocity components in the three dimensions, the total velocity vector, the updated velocity vector and the three-dimensional coordinate, i.e., the corrected three-dimensional measurement, can be obtained.

[0143] In summary, in the track initiation stage, the matching between points is converted into the matching between tracks, and in the multi-target and clutter interference scene, the track matching scheme proposed in the embodiment can greatly improve the measurement matching speed and accuracy compared with the point-to-point matching algorithm, and a more accurate three-dimensional starting track can be obtained according to at least two two-dimensional measurements detected by radars.

[0144] Next, the second stage of the embodiment is described: data association.

[0145] The track initiation stage described above obtains the three-dimensional starting track, and the data association stage is a stage after the three-dimensional starting track is obtained, in which the missing-dimensional measurements (i.e., two-dimensional measurements) of different radars are fused and matched to generate three-dimensional measurement data of the track after the three-dimensional starting track.

[0146] Data association after track initiation is a core process of target tracking.

[0147] Currently, after track initiation, three-dimensional target points cannot be directly associated, and three-dimensional measurement points need to be generated by matching two-dimensional measurements first. However, there are certain difficulties in directly matching points to generate three-dimensional measurement points using two-dimensional measurements, and the direct matching is slow and low in accuracy.

[0148] In the data association stage, the embodiment converts the matching between points into the matching of gates. The conversion into the matching of gates can be understood as correcting the gates and matching points within the corrected gates.

[0149] Specifically, the predicted value of the three-dimensional track can be fed back to the single radar, the association gate is updated, the measurement is accurately associated by using the gate matching, and the three-dimensional tracking accuracy is improved by three-dimensional probability weighted tracking. The predicted value of the three-dimensional track is obtained by predicting the three-dimensional initial track.

[0150] Since the predicted value of the three-dimensional track is complete and has high reliability, the predicted value of the three-dimensional track is fed back to the dimension-lacking radar, the 2D gate of the dimension-lacking radar is corrected, the 2D tracking result is corrected, and a feedback loop of 2D to 3D to 2D is formed.

[0151] As shown in FIG. 9, the feedback loop of 2D to 3D to 2D can be that after obtaining the three-dimensional initial track, the three-dimensional initial track can be predicted to obtain a three-dimensional track prediction, the gate corresponding to the two-dimensional predicted value (2D) obtained according to the horizontal radar (i.e., the second radar) and the vertical radar (i.e., the first radar) is corrected according to the three-dimensional track prediction (3D), to obtain a corrected gate corresponding to the horizontal radar and the vertical radar, the two-dimensional measurement values falling into the corrected gate of the two radars are matched (the measurement matching shown in FIG. 10, i.e., the matching between the two-dimensional measurement values of the horizontal radar and the vertical radar), the three-dimensional track (3D) is updated, and the subsequent three-dimensional track prediction value is obtained by predicting according to the updated three-dimensional track. In this way, the feedback loop of 2D to 3D to 2D is formed. The three-dimensional track update means that a three-dimensional point or a three-dimensional track segment is taken as an update object.

[0152] Specifically, in the embodiment, as shown in FIG. 10, the data association after the track initiation can include the following steps:

[0153] S201, obtaining a three-dimensional predicted value based on a three-dimensional initial track.

[0154] The three-dimensional predicted value is a predicted value of a three-dimensional track obtained by predicting the three-dimensional initial track.

[0155] As shown in FIG. 11, after obtaining the three-dimensional initial track, the three-dimensional track prediction value can be obtained by prediction. The prediction manner can be that a Kalman filter is used to predict the next state of the three-dimensional track that has been obtained to obtain the predicted value of the three-dimensional track. The prediction manner is not limited to the Kalman filter, and can also be other prediction manners, such as an Extended Kalman filtering (EKF), an Unscented Kalman Filter (UKF), and an Ensemble Kalman filtering (EnKF).

[0156] S202, based on the two-dimensional measurements corresponding to the first radar and the second radar, respectively obtaining two-dimensional predicted values corresponding to the first radar and the second radar.

[0157] Similarly, as shown in FIG. 11, according to the two-dimensional measurement data of the first radar and the second radar, prediction is performed to obtain two-dimensional predicted values corresponding to the first radar and the second radar respectively. The prediction mode can adopt one or more of the prediction modes in S201 described above, and can be the same as or different from the prediction mode of the three-dimensional track.

[0158] S203, according to the three-dimensional predicted value, correcting the gate corresponding to the two-dimensional predicted value to obtain a corrected gate.

[0159] Based on the 2D gate correction of the three-dimensional track prediction feedback, the predicted value P thr of the Kalman filter of the three-dimensional space track can be fed back to the association process of the two dimensionally deficient radars P h and P s , and the predicted value and the association gate of the track of the two radars are corrected by using the three-dimensional predicted value.

[0160] Among them, the calculation formula is as follows:

[0161] Among them, P h represents the two-dimensional predicted value of radar h (the first radar), represents the updated two-dimensional predicted value corresponding to radar h, and a is a weight. P thr represents the three-dimensional predicted value. g represents a linear function, for example, g(P thr -P h ) = (P thr -P h ). Similarly, for the calculation formula of the correction process of the gate of radar s, please refer to the calculation formula of the correction process of radar h described above.

[0162] In the two-dimensional predicted track obtained from the two-dimensional measurement, the association gate (referred to as gate) before correction is a preset range with the predicted point as the center. According to the three-dimensional predicted value, the gate corresponding to the two-dimensional predicted value is corrected, which can be to determine a point according to the three-dimensional predicted value and the two-dimensional predicted value corresponding to the same time stamp, and the corrected gate is centered on the newly determined point.

[0163] For ease of understanding, an example of gate correction is listed below in conjunction with FIG. 12.

[0164] As shown in FIG. 12, point B (equivalent to P h in the formula above) represents a point in the two-dimensional predicted value corresponding to the first radar, and the original gate G1 centered on point B. Point A (equivalent to P thr) is a point in the three-dimensional track prediction value, then according to the point A and the point B, the following can be obtained: B' = A * a + (1-a) * B

[0165] wherein a is a weight, which can be preset, and B' (equivalent to a in the above formula) ) is a newly determined center point, and the corrected wave gate G1' is a wave gate with the point B' as the center. In this way, the original wave gate G1 is corrected and moved to G1' shown in the figure.

[0166] In this way, according to the above description, the wave gate of the two-dimensional prediction value corresponding to the second radar is corrected according to the three-dimensional track prediction value, to obtain the corrected first wave gate corresponding to the first radar and the corrected second wave gate corresponding to the second radar.

[0167] The corrected wave gate can be used to re-determine the two-dimensional measurement values falling within the wave gate. As shown in FIG. 13, due to errors, part of the two-dimensional measurements fails to fall within the wave gate, and after the wave gate correction, the three two-dimensional measurement values all fall within the corrected wave gate.

[0168] S204, based on the corrected wave gate, performing three-dimensional probability-weighted data association to obtain a measurement value of a three-dimensional track.

[0169] The 3D target tracking based on wave gate matching is to perform fusion matching of the double-radar point clouds by using the corrected associated wave gates of the two radars, to generate m*n three-dimensional measurements. Then, through joint calculation of the two-dimensional radar association probabilities, the optimal measurement point (the measurement value of the three-dimensional track obtained through data association) is calculated through probability weighting according to the m*n three-dimensional measurements (initial values).

[0170] The m points falling within the corrected first wave gate in the point cloud detected by the first radar are fusion matched with the n points falling within the corrected second wave gate in the point cloud detected by the second radar.

[0171] Specifically, for example, as shown in FIG. 14, the m points falling within the first wave gate corresponding to the first radar (radar 1) are fusion matched with the n points falling within the second wave gate corresponding to the second radar (radar 2), to obtain initial values of three-dimensional measurements.

[0172] Through the fusion matching, there are m*n matching assumptions between the m points and the n points, and the three-dimensional coordinates under each matching assumption are calculated to obtain m*n three-dimensional coordinates T kwhere k represents the kth matching hypothesis. For example, according to the two-dimensional coordinates of the ith point in the m points and the two-dimensional coordinates of the jth point in the n points, the three-dimensional coordinates under a matching hypothesis are calculated. The specific calculation method is the same as the method of calculating the three-dimensional coordinates based on the two-dimensional coordinates in the above steps, for example, the two-dimensional coordinates of the ith point can be projected into the coordinate system of another radar, or into the same coordinate system as the jth point, to obtain the missing dimension of the jth point, and then to complete the three dimensions to obtain the three-dimensional coordinates. Or, the two-dimensional coordinates of the jth point are projected into the same coordinate system as the ith point to obtain the three-dimensional coordinates.

[0173] Next, a three-dimensional tracking target point is generated by using three-dimensional probability weighting:

[0174] First, three-dimensional normalized probability calculation is performed to obtain three-dimensional point probability P k : P k = P i *P j / ∑ i＝1～m ∑ j＝1～n (P i *P j )

[0175] P i represents the probability that the ith point in the m points belongs to the current target track. P j represents the probability that the jth point in the n points belongs to the current target track. P k k in P i represents the kth matching hypothesis. P j P j is the association probability of the two-dimensional point matching corresponding to different radars.

[0176] Then, according to the three-dimensional point probability and the three-dimensional coordinates T k , the three-dimensional optimal measurement point is calculated:

[0177] where T k represents the three-dimensional coordinates under the kth matching hypothesis, P k represents the three-dimensional point probability under the kth matching hypothesis, represents the three-dimensional optimal measurement point, and the three-dimensional optimal measurement point is taken as the fused three-dimensional measurement value after matching.

[0178] For example, in FIG. 14, m=3, n=2, 3*2=6 matching hypotheses are obtained, the three-dimensional point probability P k corresponding to each matching hypothesis is calculated, and the three-dimensional coordinates (initial value) T k corresponding to each matching hypothesis is calculated, and

[0179] It should be noted that the point-to-point brute-force matching method in the related art is to match each of the M points in the point cloud of the first radar with each of the N points in the point cloud of the second radar, and in the scheme proposed in the embodiment of the present application, only two-by-two matching between the m points in the wave gate and the n points is required, m << M, n << N. Thus, through double-radar wave gate matching, the number of two-dimensional measurement matching is greatly reduced, and the matching accuracy is improved.

[0180] The third stage, trajectory completion, is described below.

[0181] When a target such as a UAV flies at a long distance in three-dimensional space, the reflection intensity changes due to the change in the angle relative to the radar, or the speed direction is tangential to the radar, and other problems, which may cause the radar to lose frames. Especially when frames are lost continuously, the three-dimensional trajectory prediction value is seriously deviated from the actual three-dimensional position, and the error needs to be continuously corrected in the tracking process.

[0182] The embodiment of the present application proposes that when at least one radar loses frames, at least one radar still has measurements, in order to ensure the continuity of the trajectory, the single-radar measurement data with missing dimensions can be used to correct and complete the three-dimensional trajectory of the target.

[0183] The specific trajectory correction and completion method is as follows:

[0184] After the track initiation stage and the data association stage, the three-dimensional track of the target is obtained.

[0185] The three-dimensional position T(r, θ, φ) of the target at the next time is predicted using the obtained global three-dimensional track.

[0186] The trust value of the generated prediction point T(r, θ, φ) is modeled, and the specific trust value calculation formula is as follows:

[0187] ScoreT * (r, θ, φ) represents the trust value of T(r, θ, φ), α t is the trust weight corresponding to each of the three dimensions, for example, α r represents the trust weight of the r dimension, α θ represents the trust weight of the θ dimension, α φ represents the trust weight of the φ dimension. α t An initial value is set at the beginning, and the value of α t is updated continuously in the process of continuously generating the track.

[0188] R d , θ d , φ d are the differences between the three-dimensional prediction value and the three-dimensional measurement value of a point in the three-dimensional track, respectively.

[0189] T(r) represents r in the three-dimensional prediction point T(r,θ,φ), T(θ) represents θ in the three-dimensional prediction point T(r,θ,φ), and T(φ) represents φ in the three-dimensional prediction point T(r,θ,φ).

[0190] Next, the three-dimensional predicted point T(r,θ,φ) is updated or corrected using the two-dimensional measurements from a single radar that did not experience frame loss. For example, the actual two-dimensional measurements P(r,θ,*) associated with the single radar that did not experience frame loss are used to obtain a more accurate three-dimensional target position T. * (r,θ,φ) corrects the global 3D trajectory.

[0191] Specifically, the target location is updated using trust values ​​and two-dimensional measurements: T * (r,θ,φ)=ScoreT * (r,θ,φ)*T(r,θ,φ)+(1-ScoreT * (r,θ,φ))*P(r,θ,*)

[0192] Among them, T * (r,θ,φ) represents the updated three-dimensional measurement value, and P(r,θ,*) represents the two-dimensional measurement value of a single radar without frame loss.

[0193] It should be noted that during the trajectory update process, the weight of the missing dimension is updated each time:

[0194] This represents the trust weight corresponding to the missing dimension after each update; α t This represents the trust weight corresponding to the missing dimension before each update; for example, in P(r,θ,*), the missing dimension is φ, and α is α. φ and S represents the trust weights in dimension φ before and after the update, respectively. new This indicates the ScoreT after each update. * (r,θ,φ),S old This indicates the ScoreT value before each update. * (r,θ,φ).

[0195] Thus, in the trajectory completion stage, based on the missing dimensions of radar without lost frames, the target's historical two-dimensional trajectory is used to correlate with the observable actual measurements. The observable actual two-dimensional measurements are then used to correct or complete the already obtained three-dimensional trajectory. In other words, multi-frame trajectory fitting is used to correct and complete the three-dimensional trajectory, avoiding large errors in three-dimensional trajectory prediction caused by lost frames or consecutive lost frames.

[0196] The above embodiments are only exemplary descriptions, and other various specific embodiments can be obtained by appropriately combining and modifying the above descriptions, and the present specification does not list them one by one.

[0197] According to the above exemplary description, the scheme proposed in the embodiments of the present application respectively proposes corresponding improvement schemes in the track initiation (such as track initiation) stage, the data association stage, and the track completion (or track correction) stage of target tracking.

[0198] Among them, in the track initiation stage, a three-dimensional target precise positioning technology based on two-dimensional track matching is proposed, which can accelerate the track initiation speed and improve the accuracy and stability of three-dimensional track initiation. Among them, the track matching based on feature weighting can improve the measurement matching speed; the three-dimensional measurement correction based on the matched track can improve the accuracy after measurement matching.

[0199] In the data association stage, a gate matching association technology based on three-dimensional track feedback is proposed, which uses the measurement accurate association of gate matching to enhance the accuracy of radar data association.

[0200] In the track completion stage, a three-dimensional track completion technology based on two-dimensional measurement guidance is proposed, which solves the problem of track breakage under the condition of single radar frame loss or obtains a three-dimensional track with large error.

[0201] It should be noted that the technologies used in the above three stages can be implemented as separate technical solutions.

[0202] For example, the target tracking method proposed in the embodiments of the present application can include a three-dimensional starting track generation method, which can be applied to a 3D target tracking scene and is generally suitable for a millimeter wave radar or a millimeter wave radar system (referred to as a millimeter wave radar). The millimeter wave radar includes at least two radars, for example, includes a first radar and a second radar, and the first radar and the second radar are complementary in the detection direction. In this scenario, the following scheme is used in the track initiation stage: performing matching of a first track and a second track to determine that the first track and the second track correspond to the same target; wherein the first track is one of at least one starting track corresponding to the first radar; the second track is one of at least one starting track corresponding to the second radar; based on a plurality of first measurement values included in the first track and a plurality of second measurement values included in the second track, a three-dimensional starting track corresponding to the same target is obtained. In this part of the embodiment, the steps after obtaining the three-dimensional starting track are not limited.

[0203] The above embodiments take the UAV as an example to specifically describe the construction of track features and the matching between tracks based on the track features. According to the exemplary description, it can be obtained that, more generally, in some embodiments, track features can be constructed, and the matching between tracks can be performed based on the track features. For example, track features of tracks corresponding to different radars are constructed, such as constructing a first feature of a first track corresponding to a first radar and constructing a second feature of a second track corresponding to a second radar. The matching between the first track and the second track can be calculating the similarity between the first feature and the second feature to determine whether the first track and the second track are tracks corresponding to the same target.

[0204] In actual applications, the first radar can detect multiple tracks of multiple targets, and therefore, the first radar can correspond to multiple first tracks, and the multiple first tracks can construct multiple first features. Similarly, the second radar can detect multiple tracks of multiple targets, and correspond to multiple second features. Then, the similarity between the first features and the second features can be calculated, that is, the multiple first features in the first radar and the multiple second features corresponding to the second radar are calculated for similarity, and if the similarity exceeds a threshold value, for example, the similarity exceeds 95%, it is considered that the two tracks corresponding to the two features come from the same target.

[0205] As described in the above embodiments, the constructed track features can include n sub-features, such as at least one of multiple features including length, position change R gap , velocity difference V gap , historical acceleration, track curvature, and the like.

[0206] After the track features are constructed, the similarity between the multiple features corresponding to the first radar and the features corresponding to the second radar is calculated. As described in the above embodiments, there are multiple methods for calculating the similarity, and in some embodiments, the similarity calculation method based on feature weighting coefficients (simply referred to as feature coefficients) proposed in the embodiments of the present application can be used:

[0207] After the first feature corresponding to the first track and the second feature corresponding to the second track are obtained, the feature coefficients corresponding to the n sub-features are determined based on the first feature and the second feature, and the similarity between the first feature and the second feature is calculated based on the feature coefficients. A feature includes n sub-features, and the feature weighting coefficients of each sub-feature are calculated through a predetermined function, and when the similarity is calculated, the feature weighting coefficients of each sub-feature are considered. The feature weighting coefficients can adjust the influence degree of different features on the similarity calculation result, so that the similarity calculation is more accurate and controllable.

[0208] In the embodiments of the present application, the missing dimension of the first radar is different from the missing dimension of the second radar. For example, in some embodiments, the first measurement value detected by the first radar includes two-dimensional coordinates missing a third dimension coordinate; the second measurement value detected by the second radar includes two-dimensional coordinates missing a first dimension or a second dimension.

[0209] After determining two trajectories (two-dimensional) corresponding to the same target through trajectory matching, it is necessary to generate a three-dimensional starting trajectory based on the two two-dimensional trajectories, in which a calculation process of obtaining a three-dimensional measurement value from two two-dimensional measurements is involved.

[0210] Specifically, assuming that the first trajectory and the second trajectory are two trajectories from or corresponding to the same target, a three-dimensional starting trajectory corresponding to the target can be obtained according to the first measurement value included in the first trajectory and the second measurement value included in the second trajectory, which can be obtained in the following manner:

[0211] For one of the points: based on the two-dimensional coordinates included in the second measurement value and the two-dimensional coordinates included in the first measurement value, the third dimension coordinate missing in the first measurement value is obtained; the two-dimensional coordinates included in the first measurement value are combined with the obtained third dimension coordinate to obtain a second three-dimensional measurement value. In this way, after obtaining the second three-dimensional measurement values of multiple points, the multiple points corresponding to the same target are connected to obtain a three-dimensional starting trajectory.

[0212] As mentioned above, the two-dimensional coordinates in the first measurement value missing a dimension compared to three-dimensional space coordinates are different from the two-dimensional coordinates in the second measurement value missing a dimension compared to three-dimensional space coordinates. For example, the first measurement value includes a first dimension coordinate and a second dimension coordinate, and is missing a third dimension; the second measurement value includes a first dimension coordinate and a third dimension coordinate, and is missing a second dimension coordinate; or the second measurement value includes a second dimension coordinate and a third dimension coordinate, and is missing a first dimension coordinate. The two-dimensional coordinates in the first measurement value are coordinates recorded based on a first coordinate system corresponding to the first radar; the two-dimensional coordinates in the second measurement value are coordinates recorded based on a second coordinate system corresponding to the second radar.

[0213] Specifically, based on the two-dimensional coordinates included in the second measurement value and the two-dimensional coordinates included in the first measurement value, the third dimension coordinate missing in the first measurement value can be obtained by conversion according to the first dimension coordinate and the second dimension coordinate included in the first measurement value, and according to the first dimension coordinate and the third dimension coordinate included in the second measurement value or according to the second dimension coordinate and the third dimension coordinate included in the second measurement value, so as to obtain the third dimension coordinate missing in the first coordinate system corresponding to the first radar. In this way, the two dimensions of the existing coordinates in the first measurement value and the obtained third dimension coordinates are combined, and a three-dimensional measurement value (second three-dimensional measurement value) of three dimensions is obtained. For example, the two-dimensional coordinates included in the first measurement value are (r, θ, *), and the two-dimensional coordinates included in the second measurement value are (r, *, φ), or the two-dimensional coordinates in the first measurement value are (*, θ, φ), and the two-dimensional coordinates in the second measurement value are (r, *, φ) or (r, θ, *), as long as the missing dimensions of the two two-dimensional coordinates are different. The missing dimension of the measurement value of each radar can be realized by controlling the direction of the beam plane formed by the antenna array of the radar, and the beam planes are perpendicular to each other, so that the dimensions are complementary. A three-dimensional measurement value is a measurement value of a point, and in this way, three-dimensional measurement values of multiple points can be obtained. By connecting adjacent points in the multiple points with a line, a trajectory can be obtained, and the obtained trajectory is a three-dimensional starting trajectory.

[0214] In a possible implementation, after obtaining the three-dimensional starting trajectory corresponding to the same target, the acceleration values corresponding to the first dimension, the second dimension and the third dimension can be calculated according to the first measurement value and the second measurement value; and the three dimensions corresponding to the speed and the position of the three-dimensional starting trajectory are respectively corrected by using the acceleration values, to obtain corrected speed information and position information.

[0215] In an example, the acceleration values corresponding to the first dimension, the second dimension and the third dimension can be obtained by averaging the first acceleration corresponding to the first radar and the second acceleration corresponding to the second radar, that is, by calculating the average of the acceleration of the first radar and the second radar in three dimensions (the average of the acceleration component) according to the first measurement value and the second measurement value. The first acceleration can be obtained according to the speed information in the first measurement value, for example, the acceleration is obtained according to the speed information in the first measurement value of the adjacent two points (for example, the two points of adjacent frames), as the first acceleration corresponding to one of the points. Similarly, the second acceleration can be obtained according to the speed information in the second measurement value. In other embodiments, the acceleration values corresponding to the first dimension, the second dimension and the third dimension can be obtained by weighted summation of the first acceleration and the second acceleration, for example, the weight corresponding to the first radar is 0.46, and the weight corresponding to the second radar is 0.54, and the like.

[0216] Alternatively, other preset functions can be used to calculate the acceleration values ​​corresponding to the three dimensions, with the first and second accelerations as independent variables. In summary, the acceleration values ​​used to correct the three-dimensional initial trajectory are calculated from the acceleration values ​​corresponding to different radars. The calculation method can be weighted summation, averaging (a type of weighted summation), or other preset functions.

[0217] By combining acceleration values ​​obtained from multiple radars to correct the three-dimensional initial trajectory, the impact of large single-radar measurement errors on the accuracy of the three-dimensional initial trajectory can be mitigated to some extent, thereby improving the accuracy of the obtained three-dimensional initial trajectory.

[0218] The target tracking method proposed in this application is an association method based on fusion matching of two-dimensional missing-dimensional measurements. It can be applied to 3D target tracking scenarios and is generally suitable for millimeter-wave radar. The millimeter-wave radar includes at least two radars, for example, a first radar and a second radar, which are complementary in detection azimuth. The method specifically includes the following steps: determining the three-dimensional predicted value corresponding to the target; determining the first predicted value corresponding to the first radar and the second predicted value corresponding to the second radar; obtaining a corrected first gate based on the three-dimensional predicted value and the first predicted value; obtaining a corrected second gate based on the three-dimensional predicted value and the second predicted value; and fusing and matching at least one first measurement value falling into the first gate and at least one second measurement value falling into the second gate to obtain a first three-dimensional measurement value.

[0219] In this system, the first and second radars are complementary in their detection orientation. Specifically, the measurement data detected by both radars are two-dimensional measurements. The dimension missing in the two-dimensional measurement detected by the first radar relative to the three-dimensional measurement is different from the dimension missing in the two-dimensional measurement detected by the second radar relative to the three-dimensional measurement. For example, the first radar may be missing a first dimension, while the second radar may be missing a second or third dimension. The first and second dimensions are different dimensions, or the first and third dimensions are different dimensions. For example, in a Cartesian coordinate system, if the first radar is missing the z-dimensional dimension, then the second radar is missing the x-dimensional or y-dimensional dimension; or, in polar coordinates, if the first radar is missing the r-dimensional dimension, then the second radar is missing the θ-dimensional or φ-dimensional dimension. If the first and second radars, through their relative positional layout design, can obtain measurement data that satisfies the above characteristics (i.e., different missing dimensions), then they are considered complementary in their detection orientation.

[0220] For example, if the beam planes formed by the antenna arrays of the first radar and the second radar are perpendicular to each other, measurement data satisfying the above characteristics can be obtained. Therefore, the first and second radars are complementary in detection azimuth, meaning that the beam planes formed by the first radar and the second radar are perpendicular to each other in three-dimensional space. Complementarity in detection azimuth includes situations where the two beam planes formed by the first and second radars are relatively perpendicular or approximately perpendicular (allowing for a certain range of error, close to 90°, such as 89.5° or 90.2°, etc.).

[0221] Based on the above exemplary description, it can be seen that, more generally, in some embodiments, determining the three-dimensional predicted value corresponding to the tracking target can be: performing prediction based on the three-dimensional initial trajectory to obtain the three-dimensional predicted value corresponding to the tracking target, for example, using a Kalman filter to perform prediction based on the three-dimensional initial trajectory to obtain the three-dimensional predicted value.

[0222] Determining the first predicted value in two dimensions corresponding to the first radar and the second predicted value in two dimensions corresponding to the second radar can be achieved by performing prediction based on the first trajectory detected by the first radar to obtain the first predicted value, and by performing prediction based on the second trajectory detected by the second radar to obtain the second predicted value. The first trajectory detected by the first radar can be understood as a two-dimensional initial trajectory determined based on the two-dimensional measurements detected by the first radar. Similarly, the second trajectory detected by the second radar can be understood as a two-dimensional initial trajectory determined based on the two-dimensional measurements detected by the second radar.

[0223] Based on the above exemplary description, when correcting a two-dimensional gate based on a three-dimensional predicted value, a corrected first gate is obtained based on the three-dimensional predicted value and a first predicted value. Specifically, this can be achieved by: obtaining a corrected first predicted value based on the three-dimensional predicted value corresponding to the same point and the two-dimensional first predicted value; and obtaining a corrected first gate based on the corrected first predicted value. Here, the first gate is a gate centered on the corrected first predicted value. Similarly, a corrected second gate can be obtained using the same method.

[0224] Referring to the above exemplary description, in some embodiments, obtaining a corrected first predicted value based on a three-dimensional predicted value and a two-dimensional first predicted value corresponding to the same point can be achieved by: calculating the corrected first predicted value based on the three-dimensional predicted value, the first predicted value, and a weighting coefficient; the weighting coefficient is used to control the proportion corresponding to the three-dimensional predicted value or the first predicted value.

[0225] Referring to the above exemplary description, the fusion matching of at least one first measurement value falling into the first gate and at least one second measurement value falling into the second gate to obtain three-dimensional measurement points can be performed as follows: determine m points falling into the first gate; determine n points falling into the second gate; and perform fusion matching on the m points and n points based on the probabilities corresponding to the m points and n points respectively to obtain the first three-dimensional measurement value.

[0226] The target tracking method proposed in this application can be a trajectory correction method. This method can be applied to 3D target tracking scenarios and is generally applicable to millimeter-wave radar or millimeter-wave radar systems (hereinafter referred to as millimeter-wave radar). The millimeter-wave radar includes at least two radars, for example, a first radar and a second radar, which are complementary in detection azimuth. For example, in some embodiments, the trajectory correction stage can adopt the following process: when the first radar does not lose frames but the second radar loses frames, determine the first two-dimensional measurement value corresponding to the first radar; correct the first three-dimensional measurement value based on the first measurement value. The first three-dimensional measurement value is obtained based on the first measurement value corresponding to the first radar and the second measurement value corresponding to the second radar. In this part of the embodiments, the steps of other stages before the trajectory correction stage are not limited, and the trajectory correction scheme proposed in this application can be compatible with other target tracking schemes. Correcting the obtained three-dimensional trajectory based on the radar that has not lost frames can reduce the impact on tracking accuracy when the radar loses frames.

[0227] It should be noted that the product obtained using the target tracking method proposed in this application has one or more of the following characteristics:

[0228] Hardware features: Detection is performed by a combination of multiple radars (at least two); this combination can be a combination of multiple independent radars, or it can be multiple independent arrays built into the same radar. That is, the radar hardware can consist of multiple independent arrays, and the antenna arrays of the different independent arrays are azimuthally complementary according to the manner defined in the above embodiments. Products possessing one or more of the above hardware features are considered to fall within the scope of protection of this application.

[0229] Interface display features (1): This application adopts a three-dimensional track initiation technology based on two-dimensional track matching. Under the condition of designing test cases where the relative distance and angle to the target do not change, the following features are displayed on the radar display interface using the technical solution of this application: the three-dimensional track is started only when there are tracks on both radars.

[0230] For example, when a UAV flies along a path that is completely tangent to the first radar at a constant angle, the distance R and angle θ acquired by the first radar remain unchanged, resulting in only a point without forming a trajectory. Finally, the target's direction changes, forming a trajectory. The second radar's trajectory remains normal throughout. After repeated experiments, the radar interface may display the following effect: Under these conditions, the radar tracking interface only forms a three-dimensional trajectory when the target's direction changes at the very end; before that, there is no initial trajectory. This is because the technical solution in this application does not use measurement point matching for initiation at the beginning of the trajectory; instead, it performs trajectory matching between tracks, thus initiating the three-dimensional trajectory only when both radars have trajectories.

[0231] Interface display features (2): This application adopts gate matching association technology based on three-dimensional trajectory feedback, and has the following features or display effects:

[0232] Design UAV clutter disturbance data. In a multi-target scenario, after the target trajectory is initiated, introduce a large number of clutter points inside the target gate and repeat the experiment. Then introduce a large number of clutter points outside the target gate and observe the radar tracking interface. If the radar cannot track correctly after introducing clutter points inside the gate, but the radar interface still displays a normal trajectory when a large number of clutter points are introduced outside the gate.

[0233] Interface display features (3): This application adopts a three-dimensional trajectory completion technology based on two-dimensional measurement guidance, which has the following features or display effects:

[0234] The design randomly interrupts measurements from some radars, causing at least one radar to drop frames. While the target trajectory is normal, the transmission signals of either radar 1 or radar 2 are randomly and intermittently interrupted multiple times. Then, the signal detection of radar 1 is turned off, and a radar 2D point with an incorrect direction is introduced to radar 2. The radar interface display effect using the technical solution of this application is as follows: Even after multiple random interruptions of a radar signal, the 3D trajectory remains continuous, and the target trajectory can be observed to be consistent with the direction of the 2D point with the incorrect direction. This is due to the application of the trajectory completion technology proposed in this application.

[0235] The aforementioned hardware features or interface display features are specific features that arise when the relevant technical solutions proposed in this application are adopted. They have a strong causal relationship. Products exhibiting one or more of the aforementioned features can be basically determined to have adopted the technical solutions proposed in this application and should fall within the scope of protection of this application.

[0236] The methods provided in the embodiments of this application can be implemented, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, they can be implemented, in whole or in part, in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DWD), or a semiconductor medium (e.g., solid-state drive). (disk, SSD, etc.). The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it; although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A target tracking method characterized by, The application is applied to a millimeter wave radar, the millimeter wave radar includes a first radar and a second radar, the first radar and the second radar are complementary in detection direction; the method comprises: determining a three-dimensional prediction value corresponding to a target; determining a first prediction value corresponding to the first radar and a second prediction value corresponding to the second radar; based on the three-dimensional prediction value and the first prediction value, a corrected first wave gate is obtained; based on the three-dimensional prediction value and the second prediction value, a corrected second wave gate is obtained; fuse and match at least one first measurement value falling into the first wave gate and at least one second measurement value falling into the second wave gate to obtain a first three-dimensional measurement value.

2. The method of claim 1, wherein, Before determining the three-dimensional prediction value corresponding to the tracked target, the method further comprises: performing matching of a first trajectory and a second trajectory to determine that the first trajectory and the second trajectory correspond to the same target; wherein the first trajectory is one of at least one starting trajectory corresponding to the first radar; the second trajectory is one of at least one starting trajectory corresponding to the second radar; based on a plurality of first measurement values included in the first trajectory and a plurality of second measurement values included in the second trajectory, a three-dimensional starting trajectory corresponding to the same target is obtained.

3. The method of claim 2, wherein performing matching of a first trajectory and a second trajectory to determine that the first trajectory and the second trajectory correspond to the same target comprises: determining a first feature corresponding to the first trajectory and a second feature corresponding to the second trajectory; determining that the first trajectory and the second trajectory correspond to the same target according to the similarity between the first feature and the second feature.

4. The method of claim 3, wherein the first feature and the second feature each include n sub-features; after determining the first feature corresponding to the first trajectory and the second feature corresponding to the second trajectory, the method further comprises: based on the first feature and the second feature, determining feature coefficients respectively corresponding to the n sub-features; based on the feature coefficients, calculating the similarity between the first feature and the second feature.

5. The method of any one of claims 2-4, wherein the first measurement value detected by the first radar includes two-dimensional coordinates lacking a third dimension coordinate; the second measurement value detected by the second radar includes two-dimensional coordinates lacking a first dimension or a second dimension; based on the first measurement value included in the first trajectory and the second measurement value included in the second trajectory, obtaining a three-dimensional starting trajectory corresponding to the same target comprises: based on the two-dimensional coordinates included in the second measurement value and the two-dimensional coordinates included in the first measurement value, obtaining the third dimension coordinate lacking in the first measurement value; combining the two-dimensional coordinates included in the first measurement value with the obtained third dimension coordinate to obtain a second three-dimensional measurement value; based on a plurality of second three-dimensional measurement values corresponding to the same target, obtaining the three-dimensional starting trajectory.

6. The method of claim 5, wherein after obtaining the three-dimensional starting trajectory corresponding to the same target, the method further comprises: According to the first measurement value and the second measurement value, acceleration values corresponding to the first dimension, the second dimension and the third dimension are calculated respectively; Using the acceleration values, three dimensions corresponding to the speed and position of the three-dimensional starting track are respectively corrected to obtain corrected speed information and position information.

7. The method of any one of claims 1-6, wherein, Based on the three-dimensional prediction value and the first prediction value, a corrected first gate is obtained, comprising: Based on the three-dimensional prediction value and the two-dimensional first prediction value corresponding to the same point, a corrected first prediction value is obtained; According to the corrected first prediction value, a corrected first gate is obtained; wherein the first gate is a gate centered on the corrected first prediction value.

8. The method of claim 7, wherein, Based on the three-dimensional prediction value and the two-dimensional first prediction value corresponding to the same point, a corrected first prediction value is obtained, comprising: According to the three-dimensional prediction value and the first prediction value, and a weighting coefficient, the corrected first prediction value is calculated; the weighting coefficient is used to control the proportion of the three-dimensional prediction value or the first prediction value.

9. The method of any one of claims 1-8, wherein, Fusing and matching at least one first measurement value falling into the first gate and at least one second measurement value falling into the second gate to obtain a three-dimensional measurement point, comprising: Determining m points falling into the first gate; Determining n points falling into the second gate; Based on the probabilities corresponding to the m points and the n points respectively, fusing and matching the m points and the n points to obtain a first three-dimensional measurement value.

10. The method of any one of claims 1-9, wherein, After obtaining the associated three-dimensional track, the method further comprises: In the case that the first radar does not lose frames and the second radar loses frames, determining the two-dimensional first measurement value corresponding to the first radar; According to the first measurement value, correcting the first three-dimensional measurement value.

11. An electronic device, comprising: The electronic device comprises: a memory for storing a computer program; a processor for executing the computer program in the memory to implement the method of any one of claims 1-10.

12. A computer storage medium, characterized in that The computer storage medium stores a computer program, and the computer program is executed by the processor to implement the method of any one of claims 1-10.

Images