A multi-target tracking method, system, device, storage medium and product

Abstract

The application discloses a multi-target tracking method, system, device, storage medium and product, which comprises the following steps: collecting first positions of each measurement point, simultaneously predicting second positions of each target at a current time; constructing a distance matrix according to the Euclidean distance between the first positions and the second positions; obtaining freshness information of each target according to the number of continuous unmatched measurement points of each target at the current time; calculating confidence information of each measurement point according to a reference signal receiving power of each measurement point and a preset penalty coefficient; constructing a penalty cost matrix according to the distance matrix, the freshness information and the confidence information; matching the measurement points with the targets according to the penalty cost matrix, and updating the latest track point of the target if the matching is successful. The application can improve the matching priority of fresh targets, reduce the matching priority of low-confidence measurement points, and further improve the correctness and accuracy of target matching.

Description

Technical Field

[0001] This invention relates to the field of integrated sensing technology, and more particularly to a multi-target tracking method, system, device, storage medium, and product. Background Technology

[0002] In the integrated communication and sensing project, the spatial information of the measurement points is detected, and the location is predicted based on the tracking time difference in order to match the measurement points with existing targets.

[0003] Existing target matching processes typically employ either the nearest neighbor algorithm or the probabilistic data interconnection algorithm. The nearest neighbor algorithm uses the measurement point closest to the target as the associated measurement point; the probabilistic data interconnection algorithm fuses the distance information of each data point before performing the matching.

[0004] Current target tracking solutions only consider the distance information between the target and the measurement point, and there is still considerable room for improvement in their matching accuracy. Summary of the Invention

[0005] The embodiments of the present invention aim to provide a multi-target tracking method, system, device, storage medium and product, which can improve the matching priority of fresh targets, while reducing the matching priority of low confidence measurement points, thereby further improving the correctness and accuracy of target matching.

[0006] In a first aspect, embodiments of the present invention provide a multi-target tracking method, comprising: Collect the first position of each measurement point, and predict the second position of each target at the current moment; Construct a distance matrix based on the Euclidean distance between the first position and the second position; The freshness information of each target is obtained based on the number of consecutive unmatched measurement points at the current time. The confidence information of each measurement point is calculated based on the reference signal received power and the preset penalty coefficient at each measurement point. Construct a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information; Based on the penalty cost matrix, the measurement points are matched with the target. If the match is successful, the latest trajectory points of the target are updated.

[0007] As an improvement to the above scheme, the step of acquiring the first position of each measurement point and simultaneously predicting the second position of each target at the current moment includes: Collect the distance, azimuth, and elevation angles of all measurement points, and calculate the first position of each measurement point; Obtain the latest timestamp of the trajectory points of each target, and predict the second position of each target at the current time based on the time difference between the latest timestamp and the current time.

[0008] As an improvement to the above scheme, the step of collecting the distance, azimuth, and elevation angles of all measurement points and calculating the first position of each measurement point includes: Based on the distance, azimuth and elevation angles of all collected measurement points, measurement points with a distance difference less than a preset distance threshold and an angle difference less than a preset angle threshold are selected to form a group of measurement points to be merged. Based on the reference signal received power of each measurement point in the group of measurement points to be merged, the measurement points in the group of measurement points to be merged are merged to obtain a list of measurement points after redundancy removal. Based on the distance measurement, azimuth angle, and elevation angle, calculate the first position of each measurement point in the measurement point list.

[0009] As an improvement to the above scheme, the step of obtaining the freshness information of each target based on the number of consecutive unmatched measurement points at the current time includes: Obtain the number of consecutive unmatched measurement points for each target at the current time, thus obtaining the number of consecutive invisible targets for each target; Based on the number of consecutive invisible times for each target, the maximum number of consecutive invisible times for all targets is obtained; Based on the number of consecutive invisible times and the maximum number of consecutive invisible times, the number of consecutive invisible times is normalized to obtain the freshness information of each target.

[0010] As an improvement to the above scheme, the step of calculating the confidence information of each measurement point based on the reference signal received power and the preset penalty coefficient at each measurement point includes: Obtain the reference signal received power at each measurement point; Based on the maximum and minimum reference signal received power of all measurement points, the reference signal received power is normalized to obtain the normalized reference signal received power of each measurement point. The confidence information of each measurement point is obtained by multiplying the preset penalty coefficient by the normalized reference signal received power.

[0011] As an improvement to the above scheme, the step of constructing the penalty cost matrix based on the distance matrix, the freshness information, and the confidence information includes: A weighted freshness matrix is ​​constructed based on the freshness information of each target, the preset matching radius, and the preset non-negative weighting factor; the dimension of the weighted freshness matrix is ​​consistent with the dimension of the distance matrix. The cost matrix is ​​constructed by adding the weighted freshness matrix to the distance matrix. Based on the confidence information, a confidence matrix is ​​constructed; the dimension of the confidence matrix is ​​the same as the dimension of the distance matrix. The cost matrix is ​​added to the confidence matrix to construct the penalty cost matrix.

[0012] As an improvement to the above scheme, the step of matching the measurement point with the target based on the penalty cost matrix, and updating the latest trajectory point of the target if the match is successful, includes: The Hungarian matching algorithm is used to match the measurement points with the target based on the penalty cost matrix; If a measurement point successfully matches the target, the latest trajectory point of the target is updated based on the measurement point, and the number of consecutive unmatched measurement points of the target is reset to zero. If no measurement point successfully matches the target, the number of consecutive unmatched measurement points for the target is updated.

[0013] Secondly, embodiments of the present invention provide a multi-target tracking system, including: The location acquisition module is used to collect the first position of each measurement point and predict the second position of each target at the current time. A distance matrix construction module is used to construct a distance matrix based on the Euclidean distance between the first position and the second position; The freshness information acquisition module is used to obtain the freshness information of each target based on the number of consecutive unmatched measurement points at the current time. The confidence information acquisition module is used to calculate the confidence information of each measurement point based on the reference signal received power and the preset penalty coefficient at each measurement point. The penalty cost matrix construction module is used to construct a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information. The target matching module is used to match the measurement points with the target based on the penalty cost matrix. If the match is successful, the latest trajectory points of the target are updated.

[0014] Thirdly, embodiments of the present invention provide a multi-target tracking device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the multi-target tracking method as described above.

[0015] Fourthly, embodiments of the present invention provide a computer-readable storage medium comprising a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to perform the multi-target tracking method as described above.

[0016] Fifthly, embodiments of the present invention provide a computer program product, the computer program product including a computer program or computer instructions, wherein when the computer program or computer instructions are executed by a processor, the multi-target tracking method described above is performed.

[0017] Compared with existing technologies, the multi-target tracking method, system, device, storage medium, and product disclosed in this invention acquire the first position of each measurement point and simultaneously predict the second position of each target at the current moment; construct a distance matrix based on the Euclidean distance between the first and second positions; obtain the freshness information of each target based on the number of consecutive unmatched measurement points at the current moment; calculate the confidence information of each measurement point based on the reference signal received power and a preset penalty coefficient; construct a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information; and perform matching between measurement points and targets based on the penalty cost matrix. If a match is successful, the latest trajectory point of the target is updated. By employing the embodiments of this invention, the matching priority of fresh targets can be improved, while the matching priority of low-confidence measurement points can be reduced, further improving the correctness and accuracy of target matching. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the steps of a multi-target tracking method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a multi-target tracking system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a multi-target tracking device provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] In the description and claims, it should be understood that the terms "first," "second," etc., used in the description and claims are only for the purpose of distinguishing the description of the same technical features, and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated, nor necessarily the order of description or chronological order. The terms are interchangeable where appropriate. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature.

[0021] In integrated communication and sensing systems, there are often scenarios where multiple nearby targets need to be tracked simultaneously, such as vehicles in dense traffic or collaborative robotic arm groups in the Industrial Internet of Things (IIoT). In the complex environment of integrated communication and sensing, the targets may be dynamic and interactive, making it unreasonable to monitor and track each target individually.

[0022] In existing technologies, measurement points that may correspond to targets are obtained through sensors, image target detection, etc. Then, the matching relationship between multiple targets and multiple measurement points is determined based on distance, thereby forming the target's trajectory and realizing target tracking.

[0023] However, in practical applications, the applicant found that existing matching algorithms, such as the nearest neighbor algorithm and the probabilistic data interconnection algorithm, all of them only use the distance information between the target and the measurement point as the basis for judgment, without considering other information about the target and the measurement point, resulting in the failure to achieve the optimal target matching accuracy.

[0024] Based on the above considerations, this invention provides a multi-target tracking method. Please refer to... Figure 1 In this embodiment, the multi-target tracking method is specifically executed through steps S1 to S6: S1. Collect the first position of each measurement point, and predict the second position of each target at the current moment; S2. Construct a distance matrix based on the Euclidean distance between the first position and the second position; S3. Based on the number of consecutive unmatched measurement points for each target at the current time, obtain the freshness information of each target; S4. Calculate the confidence information of each measurement point based on the reference signal received power and the preset penalty coefficient. S5. Construct a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information; S6. Based on the penalty cost matrix, match the measurement points with the target. If the match is successful, update the latest trajectory points of the target.

[0025] The first position is the coordinate of the measurement point at the current moment, which is the actual position of the target observed through real-time sensing. However, in the case of multiple measurement points, it is difficult to determine which target the measurement point corresponds to. Therefore, it is necessary to further perform theoretical prediction of the target's second position to obtain a matching benchmark between the measurement point and the target.

[0026] In some preferred embodiments, the second position is the position where the target is most likely to appear at the current moment, calculated based on the target's historical trajectory.

[0027] The Euclidean distance between the first and second positions reflects the spatial proximity between the prediction and the observation. The closer the first position of the measurement point is to the second position of the target, the more likely the entity corresponding to the measurement point is the target.

[0028] In some preferred embodiments, the Euclidean distance between the first position of measurement point j and the second position of target i is expressed as: (1); in, Let j be the first position of the measurement point; This represents the second position of target i.

[0029] By pairwise pairwise combinations of measurement points and the target, and calculating the Euclidean distance between the first and second positions, the final distance matrix can be obtained. In this embodiment of the invention, the distance matrix has I rows and J columns; where I is the total number of all targets and J is the total number of all measurement points.

[0030] It should be noted that in the existing technology, the distance matrix or a simply transformed distance matrix is ​​directly input into matching algorithms such as nearest neighbor algorithm and probabilistic data interconnection algorithm. During the matching process, only the distance between the current measurement position and the target prediction position is used as the basis for matching, and the influence of other factors on the accuracy of target matching is rarely considered.

[0031] In this embodiment of the invention, freshness information and confidence information are also superimposed on the distance matrix to guide the matching algorithm to make more robust decisions, rather than simply selecting the nearest measurement point.

[0032] Freshness information is derived from the number of consecutive unmatched measurement points for a target, describing the target's waiting time. If a target fails to match new data for an extended period, its freshness changes, typically meaning it may need to be removed from the matching pool or given lower priority in matching.

[0033] Reference Signal Received Power (RSRP) is an important parameter for measuring signal strength in wireless communication. The stronger the RSRP at a measurement point, the greater the likelihood that the measurement point originates from a real target, and the higher the confidence level.

[0034] The penalty cost matrix is ​​a final decision matrix that integrates distance, freshness, and confidence. The goal of the matching algorithm is to find the optimal matching solution within the penalty cost matrix. During the final decision-making process, the closer the Euclidean distance between the first and second positions, the higher the target's freshness, and the higher the confidence of the measurement point, the more likely the target and measurement point are to match.

[0035] In the above scheme, based on distance information, the matching priority of fresh targets can be improved. At the same time, the confidence information of measurement points is used as a penalty term and fused with the former two to reduce the matching priority of low-confidence measurement points, thereby improving the correctness and accuracy of target matching.

[0036] As a preferred implementation, step S1 involves acquiring the first position of each measurement point and simultaneously predicting the second position of each target at the current moment, which is executed through steps S11-S12: S11. Collect the distance, azimuth and elevation angles of all measurement points, and calculate the first position of each measurement point; S12. Obtain the latest timestamp of the trajectory points of each target, and predict the second position of each target at the current time based on the time difference between the latest timestamp and the current time.

[0037] The raw data directly collected by the sensor is usually in polar or spherical coordinates. In some preferred embodiments, in order to achieve unified calculation of subsequent Euclidean distance, the first position is the coordinate of the measurement point in a three-dimensional rectangular coordinate system.

[0038] By transforming the coordinate system, the first position of the measurement point can be calculated. For example, the first position of the measurement point... Represented as: (2); in, For distance measurement; It is the azimuth angle; It is the pitch angle.

[0039] The target trajectory points that can be obtained at the current moment are usually old data, which are out of sync with the data at the first position. Therefore, it is necessary to extrapolate the state of both the measurement point and the target to the same moment so that the two can be directly compared.

[0040] In some preferred embodiments, the time difference between the latest timestamp and the current time is input into the Kalman filter of the target, and the second position of each target at the current time is predicted by the Kalman filter.

[0041] The above scheme achieves complete alignment between the data of the measurement points and the data of the target in both time and space, providing high-quality input data pairs for subsequent matching.

[0042] Further, preferably, step S11, collecting the distance, azimuth, and elevation angles of all measurement points, and calculating the first position of each measurement point, includes: Based on the distance, azimuth and elevation angles of all collected measurement points, measurement points with a distance difference less than a preset distance threshold and an angle difference less than a preset angle threshold are selected to form a group of measurement points to be merged. Based on the reference signal received power of each measurement point in the group of measurement points to be merged, the measurement points in the group of measurement points to be merged are merged to obtain a list of measurement points after redundancy removal. Based on the distance measurement, azimuth angle, and elevation angle, calculate the first position of each measurement point in the measurement point list.

[0043] When sensors collect raw data, a real target often generates multiple similar measurement points. If these points are directly fed into the matching algorithm without processing, it will lead to a surge in computation and tracking confusion.

[0044] The preset distance threshold is the ranging error threshold for the same target, which varies depending on the target's shape, properties, and the algorithm used. The preset angle threshold is the angle measurement error threshold for the same target, which also varies depending on the target's shape, properties, and the algorithm used. When both the distance difference and the angle difference are less than the preset distance threshold, it is impossible to spatially distinguish multiple measurement points.

[0045] In this embodiment of the invention, by filtering based on two dimensions—distance difference and angle difference—a group of measurement points that are highly clustered in space is identified and subsequently merged, leaving only one measurement point in each group for the subsequent matching process. The redundancy removal benchmark is the reference signal received power of the measurement point; measurement points with higher reference signal received power generally have higher reflection reliability.

[0046] Preferably, the measurement point with the highest reference signal received power in the group of measurement points to be merged can be retained, or the weighted average position in the group of measurement points to be merged can be calculated using the reference signal received power as the weight, and this average position can be used as the retained measurement point.

[0047] In the above scheme, redundant data of the reported measurement points are deduplicated, which can significantly reduce the number of measurement points participating in the matching calculation, thereby reducing the time and space complexity of the matching calculation.

[0048] As a preferred implementation, step S3, obtaining the freshness information of each target based on the number of consecutive unmatched measurement points at the current time, includes: Obtain the number of consecutive unmatched measurement points for each target at the current time, thus obtaining the number of consecutive invisible targets for each target; Based on the number of consecutive invisible times for each target, the maximum number of consecutive invisible times for all targets is obtained; Based on the number of consecutive invisible times and the maximum number of consecutive invisible times, the number of consecutive invisible times is normalized to obtain the freshness information of each target.

[0049] After each measurement point and target matching operation, the system automatically iterates through all tracked targets, checking whether each target has successfully matched a measurement point in the current frame. If no match is found, the system increments the target's counter by 1; if a match is found, the counter is reset to zero. The value of this counter represents the number of consecutive invisible targets.

[0050] By using the maximum consecutive invisibility count, a dynamic reference can be established for the current state of all targets, and then the consecutive invisibility count of each target can be normalized. The closer the normalized consecutive invisibility count is to 0, the higher its freshness at the current moment; conversely, the closer the normalized consecutive invisibility count is to 1, the lower its freshness at the current moment.

[0051] In some preferred embodiments, the normalized consecutive invisible number of target i Represented as: (3); in, Let i be the number of consecutive times that target i is invisible. The maximum number of consecutive invisible targets.

[0052] The above scheme enables all targets to participate in the calculation with freshness on the same order of magnitude, and allows for dynamic adjustment of the memory duration for targets that have not been matched consecutively due to missed detections or other reasons, based on the overall situation of the scenario.

[0053] In a preferred embodiment, step S4, calculating the confidence information of each measurement point based on the reference signal received power and a preset penalty coefficient, includes: Obtain the reference signal received power at each measurement point; Based on the maximum and minimum reference signal received power of all measurement points, the reference signal received power is normalized to obtain the normalized reference signal received power of each measurement point. The confidence information of each measurement point is obtained by multiplying the preset penalty coefficient by the normalized reference signal received power.

[0054] In integrated sensing or radar sensing systems, the reference signal received power is a key indicator for measuring the magnitude of echo energy. A higher reference signal received power indicates a greater likelihood that the data at that measurement point originates from a reflection of a real target, and a lower degree of noise contamination.

[0055] The normalized reference signal received power removes the absolute value of the original power, retaining only the relative strength of the measurement point among all measurement points at the current moment, which can clearly represent the quality relationship between measurement points.

[0056] In some preferred embodiments, the normalized reference signal received power at measurement point j Represented as: (4); in, Let J be the reference signal received power at measurement point j; The maximum reference signal received power at all measurement points; This represents the minimum reference signal received power at all measurement points.

[0057] The penalty coefficient is used to adjust the scale or influence of confidence information in the overall matching. The penalty coefficient is an adjustable parameter. When constructing the penalty cost matrix, normalizing the reference signal received power and the penalty coefficient ensures a reduction in the matching cost of high-confidence measurement points, thereby encouraging the algorithm to prioritize these high-quality measurement points for association.

[0058] It should be noted that the penalty coefficient in this embodiment of the invention is a non-positive number. This is because the element values ​​of both the distance matrix and the freshness matrix have a negative impact on the matching cost; that is, the larger the element value in the distance matrix and the freshness matrix, the lower the matching probability between the corresponding target and the measurement point. However, for the confidence of the measurement point, the higher the normalized reference signal received power, the higher the matching probability between the corresponding target and the measurement point should be. Therefore, the penalty coefficient is used to adjust the normalized reference signal received power so that it can be applied to the penalty cost matrix for target tracking.

[0059] In the above scheme, by normalizing the received power of the reference signal at each measurement point at the current moment, the confidence information can be made to have environmental adaptability; the penalty coefficient can also introduce the confidence information into the matching dimension, thereby improving the accuracy of target tracking in cluttered environments.

[0060] As a preferred implementation, step S5, constructing a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information, includes: A weighted freshness matrix is ​​constructed based on the freshness information of each target, the preset matching radius, and the preset non-negative weighting factor; the dimension of the weighted freshness matrix is ​​consistent with the dimension of the distance matrix. The cost matrix is ​​constructed by adding the weighted freshness matrix to the distance matrix. Based on the confidence information, a confidence matrix is ​​constructed; the dimension of the confidence matrix is ​​the same as the dimension of the distance matrix. The cost matrix is ​​added to the confidence matrix to construct the penalty cost matrix.

[0061] In the above embodiments of the present invention, the dimensions of the distance matrix include the target and the measurement points. It is understood that, in order to achieve the subsequent superposition of the weighted freshness matrix and the confidence matrix on the distance matrix, it is necessary to ensure that the weighted freshness matrix and the confidence matrix have the same dimensions as the confidence matrix.

[0062] Freshness information originates from the targets; each target has corresponding freshness information, and the freshness information of all targets can form a one-dimensional array. Similarly, confidence information originates from the measurement points; each measurement point has corresponding confidence information, and the confidence information of all targets can form a one-dimensional array. However, a one-dimensional array cannot be directly added or subtracted from the distance matrix. In this embodiment of the invention, the dimensions of both are increased to obtain a weighted freshness matrix and a confidence matrix.

[0063] Each element in the distance matrix has a specific dimensional meaning in its row and column. Therefore, the directionality of dimensional expansion needs to be considered when constructing the weighted freshness matrix and confidence matrix.

[0064] In this embodiment of the invention, the elements in the distance matrix Let be the Euclidean distance between the second position of target i and the first position of measurement point j. That is, in the distance matrix, rows correspond to targets, and columns correspond to measurement points.

[0065] Since freshness information is an attribute attached to the target, it is expanded row by row when constructing the weighted freshness matrix. In this way, regardless of which measurement point the target is matched with, it will be given the same freshness cost determined by the target's own state.

[0066] Similarly, since confidence information is an attribute attached to the measurement point, the confidence matrix is ​​expanded column-wise. This way, regardless of which target the measurement point is matched with, it will be given the same confidence penalty determined by the signal strength of that measurement point.

[0067] Based on this, the weighted freshness matrix Represented as: (5); in, The pre-defined non-negative weighting factor; is the preset matching radius; I is the total number of all targets.

[0068] The matching radius is a pre-set spatial threshold used to convert dimensionless freshness information into a meaningful numerical value, which can be directly added to the Euclidean distance in the distance matrix for subsequent matching.

[0069] The non-negative weighting factor is an adjustable parameter used to control the influence of freshness information on the penalty cost matrix used for matching.

[0070] confidence matrix Represented as: (6); in, This is the preset penalty coefficient; Let J be the normalized reference signal received power at measurement point j; J is the total number of all measurement points.

[0071] Furthermore, in a preferred embodiment of the present invention, the penalty cost matrix Represented as: (7); in, It is a distance matrix; This is a weighted freshness matrix; This is the cost matrix; This is the confidence matrix.

[0072] In the above scheme, the constructed penalty cost matrix can fuse three-dimensional data of distance, target freshness, and measurement point confidence, thereby improving the correctness and accuracy of target matching.

[0073] As a preferred implementation, step S6 involves matching the measurement points with the target based on the penalty cost matrix. If the match is successful, the latest trajectory points of the target are updated, including: The Hungarian matching algorithm is used to match the measurement points with the target based on the penalty cost matrix; If a measurement point successfully matches the target, the latest trajectory point of the target is updated based on the measurement point, and the number of consecutive unmatched measurement points of the target is reset to zero. If no measurement point successfully matches the target, the number of consecutive unmatched measurement points for the target is updated.

[0074] It should be noted that the Hungarian matching algorithm is a combinatorial optimization algorithm used to solve the assignment problem, that is, how to optimally allocate N targets and M measurement points in a one-to-one manner with minimum cost. In this embodiment of the invention, the penalty cost matrix is ​​used as input, and the Hungarian algorithm searches for the globally optimal matching scheme in the penalty cost matrix.

[0075] When a target and a measurement point pair are successfully matched, it means that the target currently has real-world observation support, and the first position of the measurement point can be directly added to the target's latest trajectory point. Simultaneously, by resetting the number of consecutive unmatched measurement points for the target to zero, the target's freshness can be reset, preventing it from being mistakenly judged as removed due to historical disconnections.

[0076] When a target fails to match a measurement point, the number of consecutive unmatched measurement points for that target is increased to reduce its freshness. In the next match, its priority will be reduced, thus allowing the measurement point to be matched with targets that have fewer consecutive unmatched times, thereby improving the accuracy of the match.

[0077] In the above scheme, the Hungarian algorithm is used to make global optimal decisions, and the target state is maintained bidirectionally based on the matching results, forming a complete closed-loop feedback system that can ensure the continuity and accuracy of tracking.

[0078] The multi-target tracking method provided in this embodiment of the invention can improve the matching priority of fresh targets based on distance information. At the same time, it integrates the confidence information of measurement points as a penalty term with the former two to reduce the matching priority of low-confidence measurement points, thereby improving the correctness and accuracy of target matching.

[0079] This invention provides a multi-target tracking system. Please refer to [link / reference]. Figure 2 The multi-target tracking system includes a location acquisition module 11, a distance matrix construction module 12, a freshness information acquisition module 13, a confidence information acquisition module 14, a penalty cost matrix construction module 15, and a target matching module 16, wherein: The position acquisition module 11 is used to collect the first position of each measurement point and predict the second position of each target at the current time. The distance matrix construction module 12 is used to construct a distance matrix based on the Euclidean distance between the first position and the second position; The freshness information acquisition module 13 is used to obtain the freshness information of each target based on the number of times each target has not been matched with measurement points at the current time. The confidence information acquisition module 14 is used to calculate the confidence information of each measurement point based on the reference signal received power and the preset penalty coefficient at each measurement point. The penalty cost matrix construction module 15 is used to construct a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information. The target matching module 16 is used to match the measurement point with the target according to the penalty cost matrix. If the match is successful, the latest trajectory point of the target is updated.

[0080] In a preferred embodiment, the location acquisition module 11 includes: The first position calculation unit is used to collect the distance, azimuth and elevation angles of all measurement points and calculate the first position of each measurement point. The second position prediction unit is used to obtain the latest timestamp of the trajectory points of each target, and predict the second position of each target at the current time based on the time difference between the latest timestamp and the current time.

[0081] Further, preferably, the first position calculation unit is specifically used for: Based on the distance, azimuth and elevation angles of all collected measurement points, measurement points with a distance difference less than a preset distance threshold and an angle difference less than a preset angle threshold are selected to form a group of measurement points to be merged. Based on the reference signal received power of each measurement point in the group of measurement points to be merged, the measurement points in the group of measurement points to be merged are merged to obtain a list of measurement points after redundancy removal. Based on the distance measurement, azimuth angle, and elevation angle, calculate the first position of each measurement point in the measurement point list.

[0082] In a preferred embodiment, the freshness information acquisition module 13 is specifically used for: Obtain the number of consecutive unmatched measurement points for each target at the current time, thus obtaining the number of consecutive invisible targets for each target; Based on the number of consecutive invisible times for each target, the maximum number of consecutive invisible times for all targets is obtained; Based on the number of consecutive invisible times and the maximum number of consecutive invisible times, the number of consecutive invisible times is normalized to obtain the freshness information of each target.

[0083] In a preferred embodiment, the confidence information acquisition module 14 is specifically used for: Obtain the reference signal received power at each measurement point; Based on the maximum and minimum reference signal received power of all measurement points, the reference signal received power is normalized to obtain the normalized reference signal received power of each measurement point. The confidence information of each measurement point is obtained by multiplying the preset penalty coefficient by the normalized reference signal received power.

[0084] In a preferred embodiment, the penalty cost matrix construction module 15 is specifically used for: A weighted freshness matrix is ​​constructed based on the freshness information of each target, the preset matching radius, and the preset non-negative weighting factor; the dimension of the weighted freshness matrix is ​​consistent with the dimension of the distance matrix. The cost matrix is ​​constructed by adding the weighted freshness matrix to the distance matrix. Based on the confidence information, a confidence matrix is ​​constructed; the dimension of the confidence matrix is ​​the same as the dimension of the distance matrix. The cost matrix is ​​added to the confidence matrix to construct the penalty cost matrix.

[0085] In a preferred embodiment, the target matching module 16 is specifically used for: The Hungarian matching algorithm is used to match the measurement points with the target based on the penalty cost matrix; If a measurement point successfully matches the target, the latest trajectory point of the target is updated based on the measurement point, and the number of consecutive unmatched measurement points of the target is reset to zero. If no measurement point successfully matches the target, the number of consecutive unmatched measurement points for the target is updated.

[0086] The multi-target tracking system provided in this embodiment of the invention can improve the matching priority of fresh targets based on distance information. At the same time, it integrates the confidence information of measurement points as a penalty term with the former two to reduce the matching priority of low-confidence measurement points, thereby improving the correctness and accuracy of target matching.

[0087] Please see Figure 3 , Figure 3This is a structural block diagram of a multi-target tracking device provided in an embodiment of the present invention. The multi-target tracking device includes a processor 31, a memory 32, and a computer program stored in the memory 32 and executable on the processor 31. When the processor 31 executes the computer program, it implements the steps in the various multi-target tracking method embodiments described above, such as steps S1 to S6.

[0088] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 32 and executed by the processor 31 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the multi-target tracking device.

[0089] The multi-target tracking device may include, but is not limited to, a processor 31 and a memory 32. Those skilled in the art will understand that the schematic diagram is merely an example of a multi-target tracking device and does not constitute a limitation on the device. It may include more or fewer components than illustrated, or combine certain components, or use different components. For example, the multi-target tracking device may also include input / output devices, network access devices, buses, etc.

[0090] The processor 31 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 31 is the control center of the multi-target tracking device, connecting various parts of the device via various interfaces and lines.

[0091] The memory 32 can be used to store the computer program and / or modules. The processor 31 implements various functions of the multi-target tracking device by running or executing the computer program and / or modules stored in the memory 32 and calling the data stored in the memory 32. The memory 32 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 32 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0092] If the modules / units integrated in the multi-target tracking device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by the processor 31, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0093] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A multi-target tracking method, characterized in that, include: Collect the first position of each measurement point, and predict the second position of each target at the current moment; Construct a distance matrix based on the Euclidean distance between the first position and the second position; The freshness information of each target is obtained based on the number of consecutive unmatched measurement points at the current time. The confidence information of each measurement point is calculated based on the reference signal received power and the preset penalty coefficient at each measurement point. Construct a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information; Based on the penalty cost matrix, the measurement points are matched with the target. If the match is successful, the latest trajectory points of the target are updated.

2. The multi-target tracking method as described in claim 1, characterized in that, The process of acquiring the first position of each measurement point and simultaneously predicting the second position of each target at the current moment includes: Collect the distance, azimuth, and elevation angles of all measurement points, and calculate the first position of each measurement point; Obtain the latest timestamp of the trajectory points of each target, and predict the second position of each target at the current time based on the time difference between the latest timestamp and the current time.

3. The multi-target tracking method as described in claim 2, characterized in that, The process involves collecting distance, azimuth, and elevation angles from all measurement points and calculating the first position of each measurement point, including: Based on the distance, azimuth and elevation angles of all collected measurement points, measurement points with a distance difference less than a preset distance threshold and an angle difference less than a preset angle threshold are selected to form a group of measurement points to be merged. Based on the reference signal received power of each measurement point in the group of measurement points to be merged, the measurement points in the group of measurement points to be merged are merged to obtain a list of measurement points after redundancy removal. Based on the distance measurement, azimuth angle, and elevation angle, calculate the first position of each measurement point in the measurement point list.

4. The multi-target tracking method as described in claim 1, characterized in that, The method of obtaining freshness information for each target based on the number of consecutive unmatched measurement points at the current time includes: Obtain the number of consecutive unmatched measurement points for each target at the current time, thus obtaining the number of consecutive invisible targets for each target; Based on the number of consecutive invisible times for each target, the maximum number of consecutive invisible times for all targets is obtained; Based on the number of consecutive invisible times and the maximum number of consecutive invisible times, the number of consecutive invisible times is normalized to obtain the freshness information of each target.

5. The multi-target tracking method as described in claim 1, characterized in that, The step of calculating the confidence information of each measurement point based on the reference signal received power and a preset penalty coefficient includes: Obtain the reference signal received power at each measurement point; Based on the maximum and minimum reference signal received power of all measurement points, the reference signal received power is normalized to obtain the normalized reference signal received power of each measurement point. The confidence information of each measurement point is obtained by multiplying the preset penalty coefficient by the normalized reference signal received power.

6. The multi-target tracking method as described in claim 1, characterized in that, The step of constructing a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information includes: A weighted freshness matrix is ​​constructed based on the freshness information of each target, the preset matching radius, and the preset non-negative weighting factor; the dimension of the weighted freshness matrix is ​​consistent with the dimension of the distance matrix. The cost matrix is ​​constructed by adding the weighted freshness matrix to the distance matrix. Based on the confidence information, a confidence matrix is ​​constructed; the dimension of the confidence matrix is ​​the same as the dimension of the distance matrix. The cost matrix is ​​added to the confidence matrix to construct the penalty cost matrix.

7. The multi-target tracking method as described in claim 1, characterized in that, The step of matching measurement points with targets based on the penalty cost matrix, and updating the latest trajectory points of the target if a match is successful, includes: The Hungarian matching algorithm is used to match the measurement points with the target based on the penalty cost matrix; If a measurement point successfully matches the target, the latest trajectory point of the target is updated based on the measurement point, and the number of consecutive unmatched measurement points of the target is reset to zero. If no measurement point successfully matches the target, the number of consecutive unmatched measurement points for the target is updated.

8. A multi-target tracking system, characterized in that, include: The location acquisition module is used to collect the first position of each measurement point and predict the second position of each target at the current time. A distance matrix construction module is used to construct a distance matrix based on the Euclidean distance between the first position and the second position; The freshness information acquisition module is used to obtain the freshness information of each target based on the number of consecutive unmatched measurement points at the current time. The confidence information acquisition module is used to calculate the confidence information of each measurement point based on the reference signal received power and the preset penalty coefficient at each measurement point. The penalty cost matrix construction module is used to construct a penalty cost matrix based on the distance matrix, the freshness information, and the confidence information. The target matching module is used to match the measurement points with the target based on the penalty cost matrix. If the match is successful, the latest trajectory points of the target are updated.

9. A multi-target tracking device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the multi-target tracking method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the multi-target tracking method as described in any one of claims 1 to 7.

11. A computer program product, characterized in that, The computer program product includes a computer program or computer instructions, which, when executed by a processor, perform the multi-target tracking method as described in any one of claims 1 to 7.

Images