Object tracking method and device

Abstract

This disclosure provides an object tracking method and apparatus, applicable to the fields of artificial intelligence and autonomous driving environmental perception. The method includes: detecting target object grid cluster information and first detection box information based on preset detection rules to obtain detection results; matching the matched detection box information with first target box information in the target trajectory based on preset matching rules to obtain a first matching result, wherein the first matching result includes non-matching trajectory information; matching the non-matching trajectory information with non-matching grid cluster information to obtain a second matching result; updating the first target box based on the second matching result and preset update rules to generate a second target box; and tracking the target object using the second target box. By performing multiple matches on both successfully and unsuccessfully matched target objects, missed target objects are supplemented, improving the stability of the tracking effect and further enhancing the safety of autonomous driving.

Description

Technical Field

[0001] This disclosure relates to the fields of artificial intelligence and autonomous driving environmental perception technology, and in particular to an object tracking method and apparatus. Background Technology

[0002] Computer vision and multi-object tracking technologies play a crucial role in the field of autonomous driving. Because 3D object detection can obtain relatively accurate positional information, most current 3D multi-object tracking algorithms adopt a detection-tracking framework. This framework designs 3D object detection and multi-object tracking separately, using the results of 3D object detection as input, and obtaining the target object's ID information through a multi-object tracking algorithm.

[0003] In the process of conceiving this disclosure, the inventors discovered the following defects in the tracking and detection of target objects in related technologies: missed or false detection of target objects leads to tracking failure, posing a significant threat to driving safety; the tracking effect in related technologies largely depends on the target object detection results, and the tracking effect is very poor when the detection results are unstable or the target is lost, which cannot meet the needs of practical applications. Summary of the Invention

[0004] In view of the above problems, this disclosure provides an object tracking method, apparatus, device, storage medium, and program product.

[0005] According to a first aspect of this disclosure, an object tracking method is provided, comprising: detecting grid cluster information of a target object and first detection box information based on a preset detection rule to obtain a detection result, wherein the detection result includes matching detection box information and non-matching grid cluster information corresponding to the detection result;

[0006] Based on preset matching rules, the above-mentioned matching detection box information is matched with the first target box information in the target trajectory to obtain a first matching result, wherein the above-mentioned first matching result includes non-matching trajectory information;

[0007] The above-mentioned non-matching trajectory information is matched with the above-mentioned non-matching raster cluster information to obtain the second matching result;

[0008] Based on the second matching result and the preset update rules, the first target box is updated to generate a second target box; and

[0009] The target object is tracked using the second target bounding box.

[0010] According to embodiments of this disclosure, the method further includes: constructing a lifecycle management platform, wherein the lifecycle management platform is used to manage target object information, trajectory information, grid cluster information, detection box information, and target box information; and updating the trajectory information and target object information in the lifecycle management platform based on the tracking results.

[0011] According to embodiments of this disclosure, the above-mentioned detection of the target object grid cluster information and the first detection box information based on preset detection rules to obtain detection results includes: determining the first target distance information between the target object grid cluster information and the first detection box information; and detecting the target object grid cluster information and the first detection box information based on the target matrix corresponding to the first target distance information and the preset detection rules to generate the above-mentioned detection results.

[0012] According to an embodiment of this disclosure, determining the first target distance information between the target object grid cluster information and the first detection box information includes: generating reference detection box information corresponding to the first detection box information based on the first detection box information; and determining the first target distance information between the target object grid cluster information and the first detection box information based on the reference detection box information.

[0013] According to an embodiment of this disclosure, determining the first target distance information between the target object grid cluster information and the first detection frame information based on the reference detection frame information includes: classifying the grid clusters in the target object grid cluster information based on the reference detection frame information to generate a first grid cluster and a second grid cluster; and determining the first target distance information between the first grid cluster, the second grid cluster and the first detection frame information.

[0014] According to an embodiment of this disclosure, the above-mentioned matching of the matching detection box information with the first target box information in the target trajectory based on a preset matching rule to obtain a first matching result includes: constructing a first cost matrix based on the intersection-union ratio of the matching detection box information and the first target box information; using the first cost matrix as the independent variable of the preset matching rule to match the matching detection box information with the first target box information to generate an initial matching result; and comparing the initial matching result with a first matching threshold to generate the first matching result, wherein the first matching result includes matching pair information corresponding to the first matching result.

[0015] According to embodiments of this disclosure, the first matching result further includes matching trajectory information, and the method further includes: inputting the matching trajectory information into the lifecycle management platform; and updating the target object information in the lifecycle management platform, wherein the target object information includes the state information of the target object.

[0016] According to an embodiment of this disclosure, the above-mentioned matching of the non-matching trajectory information with the non-matching grid cluster information to obtain a second matching result includes: determining a second target distance information between the non-matching trajectory information and the non-matching grid cluster information; constructing a second cost matrix using the second target distance information; and generating the second matching result according to the preset matching rules and the second matching threshold, wherein the second matching result includes matching pair information corresponding to the second matching result.

[0017] According to embodiments of this disclosure, the second matching result includes matching pair information corresponding to the second matching result; wherein, updating the first target box based on the second matching result and a preset update rule to generate a second target box includes: generating a first coordinate value corresponding to the target dimension based on the coordinate information in a first coordinate system according to the matching pair information corresponding to the second matching result; generating a second coordinate value of the matching pair information in a second coordinate system based on the coordinate value corresponding to the target dimension; and updating the first target box using the second coordinate value and the preset update rule to obtain the second target box.

[0018] A second aspect of this disclosure provides an object tracking device, comprising:

[0019] The information detection module is used to detect the grid cluster information of the target object and the first detection box information based on the preset detection rules, and obtain the detection results. The detection results include the matching detection box information and the non-matching grid cluster information corresponding to the detection results.

[0020] The first information matching module is used to match the above-mentioned matching detection box information with the first target box information in the target trajectory based on the preset matching rules to obtain a first matching result, wherein the above-mentioned first matching result includes non-matching trajectory information;

[0021] The second information matching module is used to match the above-mentioned non-matching trajectory information with the above-mentioned non-matching raster cluster information to obtain a second matching result;

[0022] The second target bounding box generation module is used to update the first target bounding box based on the second matching result and a preset update rule, thereby generating a second target bounding box; and

[0023] The target object tracking module is used to track the target object using the second target bounding box.

[0024] A third aspect of this disclosure provides an electronic device comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors perform the method described above.

[0025] A fourth aspect of this disclosure also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the methods described above.

[0026] The fifth aspect of this disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0027] According to the object tracking method, apparatus, device, storage medium, and program product provided in this disclosure, target object grid cluster information and first detection box information are detected based on preset detection rules to obtain matching detection box information. Then, the matching detection box information and the first target box are matched based on preset matching rules to generate a first matching result. Next, the non-matching trajectory information and non-matching grid cluster information in the first matching result are matched a second time to generate a second matching result. The first target box can then be dynamically updated in real time according to the second matching result and preset update rules to obtain a second target box and track the target object. Since the target object tracking process involves multiple matchings of both successfully and unsuccessfully matched trajectories and grid clusters, missed target objects are supplemented, improving the stability of the tracking effect and further enhancing the safety of autonomous driving. Attached Figure Description

[0028] The foregoing contents, as well as other objects, features, and advantages of this disclosure, will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0029] Figure 1 The illustrations depict application scenarios of object tracking methods, apparatuses, devices, storage media, and program products according to embodiments of the present disclosure.

[0030] Figure 2 A flowchart illustrating an object tracking method according to an embodiment of the present disclosure is shown schematically.

[0031] Figure 3 A schematic diagram of a multi-level matching and lifecycle management platform according to an embodiment of the present disclosure is shown.

[0032] Figure 4The illustration shows a schematic diagram of the boundary distance calculation between the target object grid cluster information and the first detection box information according to an embodiment of the present disclosure;

[0033] Figure 5 This illustration schematically shows a second target box obtained by updating a first target box according to an embodiment of the present disclosure;

[0034] Figure 6 This illustration schematically shows a defect in a target object tracking algorithm in the related art according to an embodiment of the present disclosure;

[0035] Figure 7 A flowchart illustrating another object tracking method according to an embodiment of the present disclosure is shown schematically;

[0036] Figure 8 A schematic block diagram of an object tracking device according to an embodiment of the present disclosure is shown.

[0037] Figure 9 A block diagram schematically illustrates an electronic device suitable for implementing an object tracking method according to an embodiment of the present disclosure. Detailed Implementation

[0038] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0039] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0040] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0041] When using expressions such as "at least one of A, B, and C", they should generally be interpreted in accordance with the meaning that is commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B, and C, etc.).

[0042] In the technical solution disclosed herein, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation entry points are provided for users to choose to authorize or refuse.

[0043] The inventors discovered that in related technologies, during the detection of target objects, missed or false detections can lead to target object tracking failures, posing a significant threat to driving safety. Furthermore, the tracking performance in these technologies largely depends on the target object detection results; when the detection results are unstable or the target is lost, the tracking performance is poor and cannot meet practical application requirements. Therefore, this disclosure detects the target object's grid cluster information and the first detection box information based on preset detection rules to obtain matching detection box information. Then, based on preset matching rules, the matching detection box information and the first target box are matched to generate a first matching result. Subsequently, the non-matching trajectory information and non-matching grid cluster information in the first matching result are matched a second time to generate a second matching result. The first target box can then be dynamically updated in real time based on the second matching result and preset update rules to obtain the second target box and track the target object. Since the target object tracking process involves multiple matchings of both successfully and unsuccessfully matched trajectories and grid clusters, it supplements the missed target objects, improves the stability of the tracking performance, and further enhances the safety of autonomous driving.

[0044] Embodiments of this disclosure provide an object tracking method, apparatus, device, storage medium, and program product. The method includes: detecting target object grid cluster information and first detection box information based on preset detection rules to obtain a detection result, wherein the detection result includes matching detection box information and non-matching grid cluster information corresponding to the detection result; matching the matching detection box information with first target box information in a target trajectory based on preset matching rules to obtain a first matching result, wherein the first matching result includes non-matching trajectory information; matching the non-matching trajectory information with non-matching grid cluster information to obtain a second matching result; updating the first target box based on the second matching result and preset update rules to generate a second target box; and tracking the target object using the second target box.

[0045] Figure 1 The illustrations depict application scenarios of object tracking methods, apparatuses, devices, storage media, and program products according to embodiments of the present disclosure.

[0046] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, the server 105, and the sensor 106. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0047] Sensor 106 may include lidar sensors, cameras and image sensors, ultrasonic sensors, and millimeter-wave radar sensors, as long as they meet the requirements of autonomous driving; the specific sensor type is not limited here. Sensor 106 can be connected to server 105 via network 104. By scanning the surrounding environment, sensor 106 can detect and identify objects such as roads, vehicles, pedestrians, and obstacles, thereby assisting the vehicle in real-time environmental perception and obstacle detection.

[0048] Users can interact with server 105 via network 104 using at least one of the first terminal device 101, second terminal device 102, and third terminal device 103 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, second terminal device 102, and third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0049] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0050] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0051] It should be noted that the object tracking method provided in this embodiment can generally be executed by server 105. Correspondingly, the object tracking device provided in this embodiment can generally be located in server 105. The object tracking method provided in this embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the object tracking device provided in this embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0052] It should be understood that Figure 1 The number of terminal devices, networks, servers, and sensors shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, sensors, and servers can be included.

[0053] Figure 2 A flowchart illustrating an object tracking method according to an embodiment of the present disclosure is shown schematically.

[0054] like Figure 2 As shown, the object tracking in this embodiment includes operations S210 to S250.

[0055] In operation S210, the target object grid cluster information and the first detection box information are detected based on the preset detection rules to obtain the detection result. The detection result includes the matching detection box information and the non-matching grid cluster information corresponding to the detection result.

[0056] According to embodiments of this disclosure, the target object grid cluster information can be obtained by a sensor, and the first detection box information can be related to the detection boxes formed based on the target object. The preset detection rule can be a target object association rule based on boundary distance measurement, which may include: calculating the boundary distance from the target object grid cluster to the detection box, using the boundary distance value as a cost matrix, and using the cost matrix as a matching cost to apply a matching algorithm (Hungarian matching method) to obtain the detection matching result between the target object grid cluster and the first detection box. The matching result may include matched (successfully matched) detection box information, unmatched (failed matching) detection box information, matched grid cluster information, and unmatched grid cluster information. It is understood that the object tracking method in this disclosure embodiment can be applied to situations where multiple target objects are tracked simultaneously, and specific limitations are not specified here.

[0057] In operation S220, the matching detection box information is matched with the first target box information in the target trajectory based on the preset matching rules to obtain the first matching result, wherein the first matching result includes non-matching trajectory information.

[0058] According to embodiments of this disclosure, the target trajectory can be the trajectory of a certain frame obtained by a sensor (e.g., the trajectory of the Kth frame), and the first target bounding box can be a target bounding box in the target trajectory. The preset matching rule can be the matching rule between the matching detection box information obtained from the above detection results and the first target bounding box. The first matching result can include non-matching (matching failed) trajectory information and matching (matching successfully) trajectory information. It is understood that the target bounding box or detection box in this disclosure can include a 3D bounding box corresponding to a 3D target object.

[0059] In operation S230, the non-matching trajectory information is matched with the non-matching raster cluster information to obtain the second matching result.

[0060] According to embodiments of this disclosure, considering the possibility of missed detection of target objects in real-world object tracking scenarios, which can easily lead to tracking failures of subsequent target objects, a second matching can be performed between the non-matching (failed matching) grid cluster information in the aforementioned detection results and the non-matching (failed matching) trajectory information in the first matching results to obtain a second matching result. The matching method may include calculating the boundary distance between the non-matching trajectory and the non-matching grid cluster, constructing a cost matrix based on this boundary distance, and using a relevant matching algorithm (e.g., Hungarian matching) to match the non-matching trajectory information and the non-matching grid cluster information, deleting matching pairs with excessively high matching costs to obtain the matching result. It is understood that the second matching result may include both successful second matching results between the non-matching trajectory information and the non-matching grid cluster information, and unsuccessful second matching results.

[0061] In operation S240, the first target box is updated based on the second matching result and the preset update rule to generate the second target box.

[0062] According to embodiments of this disclosure, for non-matching trajectory information and non-matching grid cluster information that have been successfully matched twice, the first target detection box can be updated using preset update rules to generate a second target box with an updated state.

[0063] In operation S250, the target object is tracked using the second target bounding box.

[0064] In one feasible embodiment, in operation S210, target object grid cluster information can be obtained by grid-based target object (obstacle) detection, and based on the first detection box information obtained by the detection network model, the n target object grid cluster information and m first detection box information are matched and detected using preset detection rules, so that ni target object grid clusters are successfully matched with mj first detection boxes, where n≥i≥1, m≥j≥1, n, m, i and j are all integers, and ni=mj.

[0065] According to embodiments of this disclosure, target object grid cluster information and first detection box information are detected based on preset detection rules to obtain matching detection box information. Then, the matching detection box information and the first target box are matched based on preset matching rules to generate a first matching result. Next, the non-matching trajectory information and non-matching grid cluster information in the first matching result are matched a second time to generate a second matching result. The first target box can then be dynamically updated in real time according to the second matching result and preset update rules to obtain a second target box and track the target object. Since the target object tracking process involves multiple matchings of both successfully and unsuccessfully matched trajectories and grid clusters, missed target objects are supplemented, improving the stability of the tracking effect and further enhancing the safety of autonomous driving.

[0066] According to embodiments of this disclosure, the object tracking method further includes: constructing a lifecycle management platform, wherein the lifecycle management platform is used to manage target object information, trajectory information, grid cluster information, detection box information, and target box information; and updating the trajectory information and target object information in the lifecycle management platform according to the tracking results.

[0067] According to embodiments of this disclosure, a lifecycle management platform can be used to store and update target object information, trajectory information, grid cluster information, detection box information, and target box information. After the first matching result is generated, the successfully matched pairs in the first matching result can be saved, and new trajectory information can be constructed for the unmatched pairs in the first matching result. After the second matching result is generated, the target object grid clusters and trajectory information that failed to match in the second time are associated, and the trajectory information in the lifecycle management platform is updated again.

[0068] In one feasible embodiment, the lifecycle management platform can also be divided into two levels based on multiple matches during the object tracking process: a first-level matching lifecycle management platform associated with the first matching result, and a second-level matching lifecycle management platform associated with the second matching result.

[0069] Figure 3 A schematic diagram of a multi-level matching and lifecycle management platform according to an embodiment of the present disclosure is shown.

[0070] like Figure 3 As shown, the lifecycle management platform can include a first-level matching lifecycle management platform 310 and a second-level matching lifecycle management platform 320. In the first-level matching lifecycle management platform 310, corresponding thresholds can be set for adding and deleting trajectories. For trajectories that successfully match in the first-level matching between the detection box and the first target box in the target trajectory, their detection failure count is set to 0; for detection boxes that fail to match in the target association between the detection box and the target trajectory, their detection success count is incremented by 1; for trajectories that fail to match in the first-level matching between the detection box and the first target box in the target trajectory, and for trajectories that fail to match in the second-level matching (i.e., matching non-matching trajectory information with non-matching raster cluster information), their missed detection count is incremented by 1; target trajectories with a detection success count greater than a certain threshold are output, and target trajectories with a detection failure count greater than a certain threshold are deleted.

[0071] In the secondary lifecycle management platform 320, for target association based on obstacle grid clusters and tracking trajectories (i.e., secondary matching of non-matching trajectory information and non-matching grid cluster information), to prevent the target trajectory from being incorrectly associated with a certain target object grid cluster for a long time, a secondary matching lifecycle platform can be designed. For trajectory i that is successfully matched in the target association based on the target trajectory and obstacle grid cluster, its number of successful matches Ni is incremented by 1; if the target trajectory is successfully associated with the detection box in the primary matching, Ni is set to 0; trajectories with Ni greater than a certain threshold are deleted.

[0072] According to embodiments of this disclosure, by setting up a multi-level lifecycle management platform, resources can be dynamically allocated and released based on the state of the target trajectory (e.g., lost, reappeared, stationary, etc.), which helps to optimize the utilization of computing resources and avoid wasting resources when the target is not being tracked. At the same time, by managing the lifecycle of the target trajectory, situations such as the appearance, disappearance, and reappearance of the target trajectory can be handled more effectively, thereby improving the stability and robustness of the tracking algorithm and improving the tracking effect.

[0073] According to embodiments of this disclosure, based on preset detection rules, the target object grid cluster information and the first detection box information are detected to obtain detection results, including: determining first target distance information between the target object grid cluster information and the first detection box information; and based on the target matrix corresponding to the first target distance information and preset detection rules, the target object grid cluster information and the first detection box information are detected to generate detection results.

[0074] According to embodiments of this disclosure, the first target distance information can be the boundary distance between the target object grid cluster information and the first detection box information, and a target matrix is ​​obtained through the first target distance information. The target matrix can characterize the cost or price of different tracking decisions. The target matrix can also be called the cost matrix. The cost matrix is ​​a two-dimensional matrix, where rows can represent the true state and columns can represent the predicted state. The value in each cell represents the cost or price of predicting the state as the value corresponding to the column when the true state is the value corresponding to the row. In multi-target object tracking algorithms, the cost matrix can describe the cost when a tracking hypothesis matches the true target. The preset detection rule can include the Hungarian matching method. After obtaining the cost matrix, the cost matrix can be used as the matching cost of the Hungarian matching method to generate the matching result between the target object grid cluster information and the first detection box information.

[0075] According to embodiments of this disclosure, by using the first target distance as the cost matrix and employing the Hungarian matching algorithm, multi-target objects can be tracked, and large-scale tasks and resources can be efficiently processed in a parallel computing environment, satisfying the robustness and parallelizability of information processing in target object tracking.

[0076] According to embodiments of this disclosure, determining the first target distance information between the target object grid cluster information and the first detection box information includes: generating reference detection box information corresponding to the first detection box information based on the first detection box information; and determining the first target distance information between the target object grid cluster information and the first detection box information based on the reference detection box information.

[0077] According to embodiments of this disclosure, reference detection box information can be used to calculate the distance between the target object grid cluster and the first detection box. The reference detection box is obtained by enlarging the first detection box by a corresponding multiple (e.g., 2 times). The size of the reference detection box can be determined according to the actual situation, and is not limited here.

[0078] According to embodiments of this disclosure, determining first target distance information between target object grid cluster information and first detection box information based on reference detection box information includes: classifying grid clusters in target object grid cluster information according to reference detection box information to generate a first grid cluster and a second grid cluster; and determining the first target distance information between the first grid cluster, the second grid cluster and the first detection box information.

[0079] According to embodiments of this disclosure, a first grid cluster can represent a grid cluster within a reference detection frame, and a second grid cluster can represent a grid cluster outside the reference detection frame. First target distance information can represent the distance values ​​between the grid clusters within and outside the reference detection frame and the first detection frame information.

[0080] Figure 4 The illustration shows a schematic diagram of the calculation of the boundary distance between the target object grid cluster information and the first detection box information according to an embodiment of the present disclosure.

[0081] like Figure 4 As shown, the first detection box 410 can first be enlarged by two times to obtain... Figure 4 The dashed box 420 in the middle divides a group of target object (obstacle) grid clusters into grids inside the box and grids outside the box, thereby calculating the boundary distance between the grid cluster and the first detection box 410. The specific calculation of the boundary distance is shown in the following formula (1):

[0082] D b =D in +D out (1);

[0083] Where D b D represents the boundary distance. in This represents the average distance from the grid cells inside the dashed box 420 to the first detection box. (D) out This represents the distance from the grid outside the dashed box to the first detection box.

[0084] Calculate the coordinates of the grid in the local coordinate system with the center of the first detection frame 410 as the origin, that is, the coordinates of the grid in the x-axis. r y r The coordinates in the coordinate system are calculated using the following formula (2):

[0085]

[0086] Where, x b and y b The x-coordinate represents the center point of the first detection frame in the radar coordinate system. i and y i Let θ represent the coordinates of the i-th grid in the radar coordinate system, and θ be the rotation angle of the first detection frame.

[0087] Because the boundary of the first detection box is perpendicular to x r y r Since the coordinate systems are parallel, the distance from the grid to the first detection frame 410 can be calculated directly.

[0088] For the grid cells inside the dashed box 420, the calculation method is as follows: (3)-(4)

[0089]

[0090]

[0091] Where n and m represent the number of grids inside the dashed frame and the number of grids outside the dashed frame, respectively, and l and w represent the length and width of the first detection box, respectively.

[0092] For the grid outside the dashed box 420, the specific calculation method is shown in the following formulas (5)-(6):

[0093]

[0094]

[0095] After obtaining the distance information of the first target, the cost matrix can be obtained from the first target distance. The cost matrix is ​​then used as the matching cost to perform Hungarian matching, and thus the matching result is obtained.

[0096] According to embodiments of this disclosure, matching the matching detection box information with the first target box information in the target trajectory based on a preset matching rule to obtain a first matching result includes: constructing a first cost matrix based on the intersection-union ratio of the matching detection box information and the first target box information; using the first cost matrix as the independent variable of the preset matching rule to match the matching detection box information with the first target box information to generate an initial matching result; and comparing the initial matching result with a first matching threshold to generate a first matching result, wherein the first matching result includes matching pair information corresponding to the first matching result.

[0097] According to embodiments of this disclosure, the first target box information can be target box information in a tracking trajectory of a certain frame. The first cost matrix can be constructed based on the intersection-over-union (IoU) ratio of the matched detection box information and the first target box information. The preset matching rule can be a relevant matching algorithm, such as the Hungarian matching method, the Hopcroft-Karp algorithm, iterative optimization algorithms, etc. The specific algorithm is not limited here, as long as it meets the information matching standard of this disclosure. The first matching threshold can include a threshold for the intersection-over-union (IoU) ratio of the matched detection box information and the first target box information. The IoU ratio ranges from (0,1), and the threshold for the IoU ratio can be, for example, 0.5. By setting the IoU threshold, the matching results are filtered, and matching pairs with smaller IoU thresholds are deleted. It is understood that the threshold for the IoU ratio can be determined according to the actual situation, and is not specifically limited here.

[0098] In object detection tasks, if the intersection-union ratio (IUR) is greater than a certain threshold (usually 0.5), the detection result is considered a correct detection; otherwise, it is considered an incorrect detection. The IUR ranges from 0 to 1. A higher value indicates a greater overlap between the detection result and the ground truth annotation, meaning the detection result is more accurate.

[0099] According to embodiments of this disclosure, by setting the intersection-union ratio and the intersection-union ratio threshold of the matching detection box information and the first target box information, detection results that have little overlap with the real annotation can be filtered out, thereby filtering out low-quality detection results and improving the accuracy and reliability of object detection.

[0100] According to embodiments of this disclosure, the first matching result further includes matching trajectory information, and the method further includes: inputting the matching trajectory information into a lifecycle management platform; and updating the target object information in the lifecycle management platform, wherein the target object information includes the state information of the target object.

[0101] According to embodiments of this disclosure, the matching trajectory information can be the trajectory information that was successfully matched in the first matching result, thereby inputting the successfully matched trajectory information into the lifecycle management platform and updating the state of the target object. Furthermore, the trajectory information that failed to match is matched a second time.

[0102] According to embodiments of this disclosure, matching non-matching trajectory information with non-matching grid cluster information to obtain a second matching result includes: determining a second target distance information between the non-matching trajectory information and the non-matching grid cluster information; constructing a second cost matrix using the second target distance information; and generating a second matching result according to a preset matching rule and a second matching threshold, wherein the second matching result includes matching pair information corresponding to the second matching result.

[0103] According to embodiments of this disclosure, a second target distance can be obtained by calculating the distance between the non-matching trajectory information and the non-matching grid cluster information; thereby, a cost matrix is ​​constructed based on the second target distance, and a second matching result is obtained by performing a secondary matching on the non-matching trajectory information and the non-matching grid cluster information through a corresponding matching algorithm.

[0104] According to embodiments of this disclosure, the second matching result includes matching pair information corresponding to the second matching result; wherein, updating the first target box based on the second matching result and a preset update rule to generate the second target box includes: generating a first coordinate value corresponding to the target dimension based on the coordinate information in the first coordinate system of the matching pair information corresponding to the second matching result; generating a second coordinate value of the matching pair information in the second coordinate system based on the coordinate value corresponding to the target dimension; and updating the first target box using the second coordinate value and the preset update rule to obtain the second target box.

[0105] According to embodiments of this disclosure, the first coordinate system may be a local coordinate system under a first target bounding box, and the second coordinate system may represent a sensor coordinate system (e.g., a radar sensor coordinate system). The target dimension may be the dimension in the x-axis and y-axis directions under the first coordinate system.

[0106] Figure 5 The illustration shows a schematic diagram of updating a first target box to obtain a second target box according to an embodiment of the present disclosure.

[0107] like Figure 5 As shown, the coordinates of the target object (obstacle) grid cluster in the local coordinate system of the first target box 510 can be calculated first, and then the maximum value x in the x-axis direction in the local coordinate system of the first target box 510 can be calculated. r max and minimum value x r min and the maximum value y in the y-axis direction. r max and minimum value y r min Then, the offset of the center coordinates is calculated according to the following formulas (7)-(8) to obtain the center of the second target box 520:

[0108]

[0109]

[0110] Among them, l err and w err These represent the distances that the first target box can move along its long side and short side in the radar coordinate system, respectively.

[0111] Furthermore, the position of the first target bounding box 510 can be updated, and the specific calculation method is shown in the following formulas (9)-(10):

[0112] x'=x+l err ×cos(θ)-w err ×sin(θ) (9);

[0113] y'=y+l err ×sin(θ)-w err ×cos(θ) (10);

[0114] Where x′ and y′ represent the coordinates of the center of the second target box in the radar coordinate system, and θ is the rotation angle of the first target box.

[0115] Finally, the length and width of the second target box 520 can be calculated using the following formulas (11)-(12):

[0116] l' = x r max -x r min (11);

[0117] w' = y r max -y r min (12);

[0118] Where l′ and w′ represent the length and width of the second target bounding box, respectively. A second target bounding box 520 is then generated based on these dimensions.

[0119] Figure 6 The illustration shows a defect in a target object tracking algorithm in the related art according to an embodiment of the present disclosure.

[0120] AB3DMOT (Associative-Embedding 3D Multi-Object Tracking) is an algorithm for 3D multi-object tracking. It combines associative filtering and embedding learning to track multiple target objects in a 3D environment across multiple time steps. However, in practical applications, some target objects may be missed, such as… Figure 6 As shown, in the AB3DMOT target object tracking algorithm, the vehicle with ID 262 in the second and fourth frames of segment 1 was missed. However, the object tracking method of this disclosure can accurately track it and infer the detection box, as shown in Table 1 below. Table 1 schematically shows the comparison of the detection effect between related technologies and this disclosure.

[0121] Table 1

[0122]

[0123] Figure 7 A flowchart illustrating another object tracking method according to an embodiment of this disclosure is shown schematically.

[0124] like Figure 7 As shown, target object (obstacle) detection based on a grid can generate target object grid cluster information 710, and first detection box information 720 can be obtained based on a relevant detection network model. Then, based on the first target distance information between the target object grid cluster and the first detection box and a preset detection rule, a detection result 730 is obtained, which includes matching detection box information and non-matching grid cluster information. Next, based on the preset matching rule, the matching detection box information is matched with the first target box information in the target trajectory 740 to obtain a first matching result 750, which includes non-matching trajectory information. Further, the non-matching trajectory information is matched with the non-matching grid cluster information to obtain a second matching result 760. Then, based on the second matching result 760 and a preset update rule, the first target box information is updated to generate a second target box 770. Finally, the second target box 770 is used to track the target object. Simultaneously, a lifecycle management platform 780 is used to manage the target trajectory information, target object information, and target object grid cluster information.

[0125] Based on the above object tracking method, this disclosure also provides an object tracking device. The following will be combined with... Figure 8 The device is described in detail.

[0126] Figure 8A schematic block diagram of an object tracking device according to an embodiment of the present disclosure is shown.

[0127] like Figure 8 As shown, the object tracking device in this embodiment includes an information detection module 810, a first information matching module 820, a second information matching module 830, a second target box generation module 840, and a target object tracking module 850.

[0128] The information detection module 810 is used to detect the target object grid cluster information and the first detection box information based on preset detection rules, and obtain detection results. The detection results include matching detection box information and non-matching grid cluster information corresponding to the detection results. In one embodiment, the information detection module 810 can be used to perform the operation S210 described above, which will not be repeated here.

[0129] The first information matching module 820 is used to match the matching detection box information with the first target box information in the target trajectory based on a preset matching rule to obtain a first matching result, wherein the first matching result includes non-matching trajectory information. In one embodiment, the first information matching module 820 can be used to perform the operation S220 described above, which will not be repeated here.

[0130] The second information matching module 830 is used to match non-matching trajectory information with non-matching raster cluster information to obtain a second matching result. In one embodiment, the second information matching module 830 can be used to perform the operation S230 described above, which will not be repeated here.

[0131] The second target box generation module 840 is used to update the first target box based on the second matching result and a preset update rule to generate a second target box. In one embodiment, the second target box generation module 840 can be used to perform the operation S240 described above, which will not be repeated here.

[0132] The target object tracking module 850 is used to track a target object using a second target bounding box. In one embodiment, the target object tracking module 850 can be used to perform the operation S250 described above, which will not be repeated here.

[0133] According to embodiments of this disclosure, the object tracking device comprises an information detection module 810, a first information matching module 820, a second information matching module 830, a second target box generation module 840, and a target object tracking module 850. By detecting the target object grid cluster information and the first detection box information based on preset detection rules, matching detection box information is obtained. Then, based on preset matching rules, the matching detection box information and the first target box are matched to generate a first matching result. Subsequently, the non-matching trajectory information and non-matching grid cluster information in the first matching result are matched a second time to generate a second matching result. The first target box can then be dynamically updated in real time according to the second matching result and preset update rules to obtain the second target box and track the target object. Since the target object tracking process involves multiple matchings of both successfully and unsuccessfully matched grid clusters and trajectories, missed target objects are supplemented, improving the stability of the tracking effect and further enhancing the safety of autonomous driving.

[0134] According to embodiments of this disclosure, the apparatus further includes: a lifecycle management platform construction module and a target object information update module. The lifecycle management platform construction module is used to construct a lifecycle management platform, wherein the lifecycle management platform is used to manage target object information, trajectory information, grid cluster information, detection box information, and target box information. The target object information update module is used to update the trajectory information and target object information in the lifecycle management platform based on tracking results.

[0135] According to embodiments of this disclosure, the information detection module includes: a first target distance information determination submodule and a detection result generation submodule. The first target distance information determination submodule is used to determine first target distance information between target object grid cluster information and first detection box information. The detection result generation submodule is used to detect the target object grid cluster information and the first detection box information based on a target matrix corresponding to the first target distance information and preset detection rules, and generate a detection result.

[0136] According to embodiments of this disclosure, the first target distance information determination submodule includes: a reference detection box information generation unit and a first target distance information determination unit. The reference detection box information generation unit is used to generate reference detection box information corresponding to the first detection box information based on the first detection box information. The first target distance information determination unit is used to determine the first target distance information between the target object grid cluster information and the first detection box information based on the reference detection box information.

[0137] According to embodiments of this disclosure, a first target distance information determination unit includes: a grid cluster generation subunit and a first target distance information determination subunit. The grid cluster generation subunit is used to classify grid clusters in the target object grid cluster information based on reference detection box information, generating a first grid cluster and a second grid cluster. The first target distance information determination subunit is used to determine first target distance information between the first grid cluster, the second grid cluster, and the first detection box information.

[0138] According to embodiments of this disclosure, the first information matching module includes: a first cost matrix construction submodule, an initial matching result generation submodule, and a first matching result generation submodule. The first cost matrix construction submodule is used to construct a first cost matrix based on the intersection-union ratio (IUU) of the matching detection box information and the first target box information. The initial matching result generation submodule is used to use the first cost matrix as an independent variable of a preset matching rule to match the matching detection box information and the first target box information, generating an initial matching result. The first matching result generation submodule is used to compare the initial matching result with a first matching threshold to generate a first matching result, wherein the first matching result includes matching pair information corresponding to the first matching result.

[0139] According to embodiments of this disclosure, the first matching result further includes matching trajectory information, and the apparatus further includes: a matching trajectory information input module and a target object information update module. The matching trajectory information input module is used to input the matching trajectory information into the lifecycle management platform. The target object information update module is used to update the target object information in the lifecycle management platform, wherein the target object information includes the state information of the target object.

[0140] According to embodiments of this disclosure, the second information matching module includes: a second target distance information determination submodule, a second cost matrix construction submodule, and a second matching result generation submodule. The second target distance information determination submodule is used to determine second target distance information between non-matching trajectory information and non-matching raster cluster information. The second cost matrix construction submodule is used to construct a second cost matrix using the second target distance information. The second matching result generation submodule is used to generate a second matching result according to a preset matching rule and a second matching threshold, wherein the second matching result includes matching pair information corresponding to the second matching result.

[0141] According to embodiments of this disclosure, the second matching result includes matching pair information corresponding to the second matching result; wherein, the second target box generation module includes: a first coordinate value generation submodule, a second coordinate value generation submodule, and a first target box update submodule. The first coordinate value generation submodule is used to generate a first coordinate value corresponding to the target dimension based on the coordinate information in a first coordinate system of the matching pair information corresponding to the second matching result. The second coordinate value generation submodule is used to generate a second coordinate value in a second coordinate system of the matching pair information based on the coordinate value corresponding to the target dimension. The first target box update submodule is used to update the first target box using the second coordinate value and a preset update rule to obtain the second target box.

[0142] According to embodiments of this disclosure, any plurality of modules among the information detection module 810, the first information matching module 820, the second information matching module 830, the second target box generation module 840, and the target object tracking module 850 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules can be combined with at least part of the functionality of other modules and implemented in one module. According to embodiments of this disclosure, at least one of the information detection module 810, the first information matching module 820, the second information matching module 830, the second target box generation module 840, and the target object tracking module 850 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in a suitable combination of any of these. Alternatively, at least one of the information detection module 810, the first information matching module 820, the second information matching module 830, the second target box generation module 840, and the target object tracking module 850 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0143] Figure 9 A block diagram schematically illustrates an electronic device suitable for implementing an object tracking method according to an embodiment of the present disclosure.

[0144] like Figure 9As shown, an electronic device 900 according to an embodiment of the present disclosure includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage portion 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0145] RAM 903 stores various programs and data required for the operation of electronic device 900. Processor 901, ROM 902, and RAM 903 are interconnected via bus 904. Processor 901 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 902 and / or RAM 903. It should be noted that the programs may also be stored in one or more memories other than ROM 902 and RAM 903. Processor 901 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0146] According to embodiments of this disclosure, the electronic device 900 may further include an input / output (I / O) interface 905, which is also connected to a bus 904. The electronic device 900 may also include one or more of the following components connected to the input / output (I / O) interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the input / output (I / O) interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 910 as needed so that computer programs read from it can be installed into the storage section 908 as needed.

[0147] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0148] According to embodiments of this disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this disclosure, the computer-readable storage medium may include ROM 902 and / or RAM 903 and / or one or more memories other than ROM 902 and RAM 903 described above.

[0149] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the object tracking method provided in the embodiments of this disclosure.

[0150] When the computer program is executed by the processor 901, it performs the functions defined in the system / apparatus of this disclosure embodiments. According to embodiments of this disclosure, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0151] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via the communication section 909, and / or installed from a removable medium 911. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0152] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, it performs the functions defined in the system of this disclosure embodiment. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0153] According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can execute entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0154] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0155] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0156] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of this disclosure is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. An object tracking method, comprising: Based on preset detection rules, the target object grid cluster information obtained by the sensor and the first detection box information obtained based on the detection network model are used for detection to obtain the detection results, including: Based on the reference detection box information with an enlarged size corresponding to the first detection box information, the grid clusters in the target object grid cluster information are classified to generate a first grid cluster within the reference detection box and a second grid cluster outside the reference detection box. Determine the first target distance information between the first grid cluster, the second grid cluster, and the first detection box information; Based on the cost matrix corresponding to the first target distance information and the preset detection rules, the target object grid cluster information and the first detection box information are detected to generate the detection result, wherein the detection result includes matching detection box information and non-matching grid cluster information corresponding to the detection result; The matching detection box information is matched with the first target box information in the target trajectory based on the preset matching rules to obtain a first matching result, wherein the first matching result includes non-matching trajectory information; The non-matching trajectory information is matched with the non-matching raster cluster information to obtain a second matching result; The first target box is updated based on the second matching result and the preset update rule to generate a second target box; and The target object is tracked using the second target bounding box.

2. The method according to claim 1, further comprising: A lifecycle management platform is constructed, wherein the lifecycle management platform is used to manage target object information, trajectory information, grid cluster information, detection box information, and target box information; and The trajectory information and target object information in the lifecycle management platform are updated based on the tracking results.

3. The method according to claim 1, wherein, The step of matching the detection box information with the first target box information in the target trajectory based on a preset matching rule to obtain a first matching result includes: Based on the intersection-union ratio of the matched detection box information and the first target box information, a first cost matrix is ​​constructed; Using the first cost matrix as the independent variable of the preset matching rule, the matching detection box information is matched with the first target box information to generate an initial matching result; and The initial matching result is compared with a first matching threshold to generate the first matching result, wherein the first matching result includes matching pair information corresponding to the first matching result.

4. The method according to claim 2, wherein the first matching result further includes matching trajectory information, and the method further includes: The matching trajectory information is input into the lifecycle management platform; as well as Update the target object information in the lifecycle management platform, wherein the target object information includes the state information of the target object.

5. The method according to claim 1, wherein, The step of matching the non-matching trajectory information with the non-matching raster cluster information to obtain a second matching result includes: Determine the second target distance information between the non-matching trajectory information and the non-matching grid cluster information; Construct a second cost matrix using the second target distance information; and The second matching result is generated according to the preset matching rule and the second matching threshold, wherein the second matching result includes matching pair information corresponding to the second matching result.

6. The method according to claim 1, wherein, The second matching result includes matching pair information corresponding to the second matching result; The step of updating the first target box based on the second matching result and a preset update rule to generate a second target box includes: Based on the coordinate information of the matching pair information corresponding to the second matching result in the first coordinate system, a first coordinate value corresponding to the target dimension is generated; Based on the coordinate values ​​corresponding to the target dimension, generate the second coordinate values ​​of the matching pair information in the second coordinate system; and The first target box is updated using the second coordinate value and the preset update rule to obtain the second target box.

7. An object tracking device, comprising: An information detection module is used to detect target object grid cluster information obtained by sensors and first detection box information obtained by a detection network model based on preset detection rules, and obtain detection results. The detection results include: classifying grid clusters in the target object grid cluster information according to a reference detection box information with an enlarged size corresponding to the first detection box information, generating a first grid cluster within the reference detection box and a second grid cluster outside the reference detection box; determining first target distance information between the first grid cluster, the second grid cluster, and the first detection box information; and detecting the target object grid cluster information and the first detection box information based on a cost matrix corresponding to the first target distance information and the preset detection rules, generating the detection results. The detection results include matching detection box information and non-matching grid cluster information corresponding to the detection results. The first information matching module is used to match the matching detection box information with the first target box information in the target trajectory based on a preset matching rule to obtain a first matching result, wherein the first matching result includes non-matching trajectory information; The second information matching module is used to match the non-matching trajectory information with the non-matching raster cluster information to obtain a second matching result; The second target box generation module is used to update the first target box based on the second matching result and a preset update rule to generate a second target box; and The target object tracking module is used to track the target object using the second target bounding box.

Images