A vehicle-road cooperation online tracking object id maintaining method and vehicle-road cooperation system

Abstract

An ID maintaining method for online tracking objects by vehicle-road cooperation and a vehicle-road cooperation system, wherein the method comprises: collecting sensing object data by a plurality of roadside detectors and storing the sensing object data in corresponding regional caches respectively; obtaining sensing object data in a plurality of adjacent regional caches and reassigning each obtained sensing object data an ID with unique identification; performing deduplication processing on the obtained sensing object data and storing the deduplicated sensing object data as a sensing object data set; and performing data association calculation on the latest obtained sensing object data and historical sensing object data in a plurality of sensing object data sets and performing ID correction on the latest obtained sensing object data according to a calculation result. The ID switch phenomenon is alleviated by performing deduplication processing on sensing object data before data association calculation and performing association calculation on a plurality of regions simultaneously.

Description

Technical Field

[0001] This invention relates to the field of multi-target tracking technology, and in particular to a method for maintaining the ID of an online tracking object in a vehicle-road cooperative system and a vehicle-road cooperative system. Background Technology

[0002] Currently, vehicle-to-everything (V2X) multi-target tracking technology is divided into two directions: offline tracking and online tracking. Offline tracking has advantages such as large data caching capacity, ample time, and the ability to perform various data processing operations, thus achieving globally optimal results. However, its real-time performance is poor. Conversely, online multi-target tracking, due to the need to ensure real-time performance, has significantly less data caching capacity and processing time compared to offline methods. This leads to frequent ID switching (i.e., frequent ID switching of the tracked object) due to issues like target occlusion, missed detections, and false detections during online data processing, resulting in unsatisfactory tracking performance and poor visibility. Furthermore, since most of the aforementioned data is roadside perception data collected simultaneously by multiple roadside detectors, overlapping areas between these detectors can cause duplicate detection of the tracked object, another reason why online tracking is prone to ID switching.

[0003] Therefore, there is an urgent need in this field for an ID retention method for vehicle-road cooperative online tracking objects that can both guarantee real-time performance and reduce ID switching phenomena. Summary of the Invention

[0004] To ensure real-time performance while reducing ID switching, in a first aspect of this invention, a method for maintaining the ID of a vehicle-road cooperative online tracking object is proposed. The method includes: collecting sensing object data from multiple roadside detectors and storing it in corresponding area caches; acquiring sensing object data from multiple adjacent area caches and reassigning a unique ID to each acquired sensing object data; deduplicating the acquired sensing object data and fusing and storing the deduplicated sensing object data into a sensing object data set; performing data association calculations between the latest acquired sensing object data and historical sensing object data in the multiple sensing object data sets, and correcting the ID of the latest acquired sensing object data based on the calculation results.

[0005] In one or more embodiments, the roadside detector includes a detection radar and / or a camera, and the perceived object data includes the following information: a circumscribed rectangle, location information, and heading angle.

[0006] In one or more embodiments, the step of reassigning a uniquely identifiable ID to each acquired sensing object data includes: reassigning a uniquely identifiable ID based on the original ID of each sensing object data; initializing ID attributes, the ID attributes including the region where the sensing object data is located, the category to which the sensing object data belongs, and unique identification information; wherein, the region where the sensing object data is located is determined by the location information of the sensing object data, the category to which the sensing object data belongs is determined by the bounding rectangle of the sensing object data, the category to which the sensing object data belongs includes designated vehicles and pedestrians, and the unique identification information includes the original ID of the sensing object data and / or license plate information.

[0007] In one or more embodiments, the deduplication process for the acquired sensing object data includes: determining whether the sensing object data is located in a detection overlap region based on the location information of the sensing object data; performing position matching on sensing object data located in the same detection overlap region one by one; and fusing the ID attributes of the two sensing object data in response to the position deviation of the two sensing object data being less than a preset value, and deleting the sensing object data with the larger ID from the corresponding region cache.

[0008] In one or more embodiments, the step of performing position matching on the sensing object data within the detection overlap area includes: performing position information matching on the sensing object data with the same classification within the detection overlap area.

[0009] In one or more embodiments, the step of performing data association calculations on the latest obtained sensing object data and historical sensing object data in multiple sensing object data sets, and correcting the ID of the latest obtained sensing object data according to the calculation results, includes: caching multiple sets of sensing object data after deduplication processing according to a preset cache capacity, wherein the sensing object data in the multiple sensing object data sets comes from detection and sensing at different times; re-dividing the cached sensing object data according to ID, assigning sensing object data with the same ID to a natural matching set, and dividing the remaining sensing object data into a soon-to-disappear set and a newly appearing set according to the order of caching; performing IOU calculations on the sensing object data in the soon-to-disappear set and the newly appearing set one by one; and correcting the ID of the newly obtained sensing object data to the ID of the corresponding historical sensing object data if the IOU calculation result between the newly obtained sensing object data in the newly appearing set and the historical sensing object data in the soon-to-disappear set is greater than a preset threshold.

[0010] In one or more embodiments, the method for maintaining the ID of a vehicle-road cooperative online tracking object according to the present invention further includes: deleting the first stored set of sensing object data when the cached set of sensing object data reaches the preset cache capacity.

[0011] In one or more embodiments, the IOU calculation includes: performing IOU calculation based on the position information, heading angle, and overlapping area of ​​the circumscribed rectangle of the newly acquired sensing object data and the historical sensing object data.

[0012] In one or more embodiments, the ID maintenance method for a vehicle-road cooperative online tracking object of the present invention further includes determining whether the sensing object data is a connected vehicle based on the unique identification information before performing data association calculation; acquiring the connected data of the sensing object data when the sensing object data is a connected vehicle; and updating the original position information and heading angle of the sensing object data based on the position information and heading angle in the connected data.

[0013] In a second aspect of the invention, a vehicle-road cooperative system is proposed, comprising: a plurality of roadside detectors configured to periodically detect and perceive a designated area; and a server including at least one processor and a memory storing an executable computer program, which, when executed by the at least one processor, is used to implement the steps of an ID maintenance method for a vehicle-road cooperative online tracking object as described in any of the above embodiments.

[0014] The beneficial effects of this invention include: by performing deduplication of the perceived object data before data association calculation and performing association calculation on multiple regions simultaneously, this invention can effectively alleviate the ID switch phenomenon; and during data association calculation, this invention filters the naturally matching perceived object data based on ID, which greatly reduces the amount of computation for data association and helps to ensure the real-time performance of online object tracking. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the ID retention method for vehicle-road cooperative online tracking objects according to the present invention.

[0017] Figure 2This is a schematic diagram illustrating a usage scenario of the ID retention method for vehicle-road cooperative online tracking objects according to the present invention;

[0018] Figure 3 This is a logic diagram for the present invention to perform data association calculations on the sensor object data in the sensor object data set obtained at different times;

[0019] Figure 4 This is a schematic diagram of a vehicle-road cooperative system according to the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to specific examples and the accompanying drawings.

[0021] It should be noted that all uses of "first" and "second" in the embodiments of the present invention are for the purpose of distinguishing two entities or parameters with the same name but different names. It is clear that "first" and "second" are only for the convenience of expression and should not be construed as limiting the embodiments of the present invention. Subsequent embodiments will not explain this in detail.

[0022] Figure 1 This is a flowchart illustrating the ID retention method for vehicle-road cooperative online tracking objects according to the present invention. Figure 1 The workflow of the ID retention method for vehicle-road cooperative online tracking objects of the present invention includes:

[0023] Step S1: Multiple roadside detectors collect data on the sensed objects and store them in their respective area caches;

[0024] Specifically, in vehicle-road cooperative systems, various roadside detectors are used to periodically collect roadside perception data in designated areas and upload the extracted object data to a backend server. The backend server then performs multi-target tracking based on real-time analysis of the object data. Commonly used roadside detectors include radar and cameras. The object data extracted by the detectors includes information such as the circumscribed rectangle, position information, and heading angle. This paper will use radar as an example to describe the method of this invention in detail.

[0025] Figure 2 This is a schematic diagram illustrating a use case of the ID retention method for vehicle-road cooperative online tracking objects according to the present invention. Figure 2As shown, detection radars for sensing are set up in areas 1, 2, and 3 respectively. The detection radars will periodically scan the detection area and upload the sensing object data. In order to achieve continuous sensing object data tracking, when setting up detection radars, the detection areas of detection radars in adjacent areas will have a certain degree of overlap, that is, there will be overlapping areas. If the comparison is not processed, it will lead to duplicate sensing object data, resulting in frequent ID switch phenomena. In addition, duplicate sensing object data will occupy a part of the area cache space, which will affect the total amount of sensing object data that the area cache can store, thus negatively affecting the success of sensing object data tracking. For example, when a certain cache area cannot store new sensing object data because duplicate cached objects occupy a large amount of space, it will negatively affect the success of subsequent object tracking.

[0026] Step S2: Obtain the sensor object data from multiple adjacent region caches, and reassign a unique ID to each obtained sensor object data;

[0027] Specifically, steps S2 to S4 will be executed by the backend server. In one embodiment, the backend server will retrieve sensing object data from multiple area caches in parallel to simultaneously obtain the latest stored sensing object data from multiple area caches. A uniquely identifiable ID is reassigned to each retrieved sensing object data, including: reassigning a uniquely identifiable ID based on the original ID of each sensing object data; initializing ID attributes, which include the area where the sensing object data is located, the category to which the sensing object data belongs, and unique identification information; wherein, the area where the sensing object data is located is determined by the location information of the sensing object data, the category to which the sensing object data belongs is determined by the bounding rectangle of the sensing object data, and the category to which the sensing object data belongs includes designated vehicles and pedestrians; the unique identification information includes the original ID of the sensing object data and / or license plate information. The purpose of this invention in reassigning IDs to each sensing object data is for subsequent tracking of sensing object data based on the new IDs and to protect the original IDs. After assigning a new ID to the sensing object data, this invention further extracts attributes from the original information of the sensing object data. For example, as mentioned above, the location information of the sensing object data determines its region, and the bounding rectangle of the sensing object data determines its category. The ID attribute facilitates the deduplication and data association calculation steps, reducing the computational load. For instance, when determining whether two sensing object data belong to the same entity through data association calculation, the calculation can be performed only on sensing object data belonging to the same category, significantly reducing computational load and ensuring real-time performance. More specifically, the unique identification information mentioned above can include information such as license plate numbers captured by a camera and the original ID information of the sensing object data. The specified vehicle can be specific to a particular vehicle type, such as a bus, public transport vehicle, van, sedan, and / or (electric) bicycle.

[0028] In this embodiment, in order to achieve a wider range of sensing object data tracking, the backend server of this invention will extract sensing object data from multiple adjacent region caches in parallel. Since there is duplicate sensing object data in the overlapping areas of adjacent regions, this invention will further perform the sensing object data deduplication process in step S3;

[0029] Step S3: Deduplicate the acquired sensing object data and merge and store the deduplicated sensing object data into a sensing object data set;

[0030] Specifically, the deduplication process for sensing object data in this invention includes: determining whether the sensing object data is located in the detection overlap area based on its location information; performing position matching on sensing object data within the same detection overlap area; and fusing the ID attributes of the two sensing object data in response to the position deviation between them being less than a preset value, and deleting the sensing object data with the larger ID from the corresponding area cache. Since each area is assigned a unique ID, a larger ID indicates more sensing object data in that area, resulting in more cached data in the corresponding area. Deleting the sensing object data corresponding to the one with the larger ID helps release the cache.

[0031] In a further embodiment, the aforementioned deduplication processing of sensing object data can also be combined with the ID attribute of the sensing object data to achieve data filtering. That is, in the process of matching the location of sensing object data within the same detection overlap area, the category to which the sensing object data belongs is first determined based on the ID attribute; then, the location information of sensing object data with the same category within the same detection overlap area is matched one by one. Deduplication processing not only avoids the problem of duplicate sensing object data and reduces ID switching, but also allows more diverse sensing object data to be stored in the region cache, providing more data support for subsequent sensing object data tracking. Specific embodiments are described below:

[0032] Please continue reading Figure 2 Suppose that there are 5 sensing object data points (A, B, C, D, and E) in regions 1, 2, and 3 respectively, where sensing object data points B and D are located in the overlapping region. At this point, 2, 3, and 2 sensing object data points can be extracted sequentially from the 3 regions, and each is assigned one of 7 IDs: IDA1, IDB1, IDB2, IDC1, IDD1, IDD2, and IDE1. These 7 sensing object data points and their IDs form a sensing object data set. Sensing object data points B and D are simultaneously detected by two detectors because they are in the overlapping detection region, thus generating two separate sensing object data points. However, these two sensing object data points (such as IDB1 and IDB2) are essentially the same sensing object data point, except for their IDs, so deduplication is required. After deduplication in step S3, a deduplicated sensing object data set is obtained, which includes 5 sensing object data points: IDA1, IDB1, IDC1, IDD1, and IDE1.

[0033] Suppose that at the next moment, the sensing object data B has completely moved into region 2 (not in any overlapping region). At this time, the backend server will be able to extract 1, 3, and 2 sensing object data from the 3 regions in sequence, and assign them to IDA1, IDB3, IDC1, IDD1, IDD2, and IDE1, respectively, for a total of 6 sensing object data. After deduplication, a set of sensing object data including IDA1, IDB3, IDC1, IDD1, and IDE1 will be obtained.

[0034] Then, step S4 can be executed to perform data association calculations on the latest obtained sensory object data and historical sensory object data in multiple sensory object data sets, and to correct the ID of the latest obtained sensory object data based on the calculation results.

[0035] Specifically, multiple sets of deduplicated sensor object data are cached according to a preset cache capacity. The sensor object data in these multiple sets originates from detections at different times. It should be noted that since the sensor object data in each set comes from sensor object data detected at the same time, each sensor object data in the same set is a unique physical individual after deduplication. The purpose of data association calculation is to track the same physical individual (the same sensor object data) at different times, so it is necessary to cache multiple sets of sensor object data from multiple different detection times before this can be performed. Then, the cached sensor object data is re-divided according to ID. Sensor object data with the same ID are assigned to the natural matching set, and the remaining sensor object data is divided into a set about to disappear and a set about to appear, based on the order of caching. IOU calculation is performed on the sensor object data in the set about to disappear and the set about to appear. In response to the IOU calculation result between newly acquired sensor object data in the set about to appear and historical sensor object data in the set about to disappear being greater than a preset threshold, the ID of the newly acquired sensor object data is corrected to the ID of the corresponding historical sensor object data. The data association calculation method used in this invention is IOU (Intersection over Union) calculation, which is a detection evaluation function calculation. Its principle is to measure the separability between two objects by the ratio of the intersection area to their union area. Before calculating the intersection or union area of ​​two rectangles, the center position of the rectangles needs to be determined using location information, and the orientation of the rectangles needs to be determined based on the heading angle. The logical diagram of this invention for data association calculation of cached perceived object data is as follows: Figure 3 As shown,

[0036] Please see Figure 3The data association algorithm primarily identifies the set A that will appear soon, the set M that will naturally match, and the set D that will soon disappear from the cached historical data set H and the data set P to be processed. Here, the historical data set H and the data set P to be processed can be understood as multiple sets of sensory objects cached before and after. If the last cached set of sensory objects is considered as the data set P to be processed, then the previously cached sets of sensory objects can be uniformly considered as the historical data set H. Here, M is the intersection of the two sets, M = H ∩ P; A is the difference between the data set to be processed and the historical data set, A = PH; and D is the difference between the historical data set and the data set to be processed, D = HP. The core of data association is to use the perceived object data in set A as a benchmark and find its best matching object in set D. The matching logic is to calculate the IOU (Intersection over Union) between Ai and Dj. IOU is defined as the ratio of the intersection area of ​​two bounding boxes to the union area of ​​their bounding boxes. Ai is the bounding rectangle of perceived object data i in A, and Dj is the bounding rectangle of perceived object data j in D. When the IOU value is greater than a threshold, the best matching perceived object data j of perceived object data i in A is obtained in D. The ID of perceived object data i is then corrected to the ID of perceived object data j, i.e., IDi = IDJ, thus completing the data association algorithm. For example, in matching motor vehicles, by calculating the IOU between the bounding rectangles of motor vehicles, when the IOU output is greater than a set threshold, it is considered that the best matching object of the motor vehicle has been found, and then the perceived object data ID of the motor vehicle is corrected.

[0037] As can be seen from the previous embodiment, the sensing object data set of the present invention contains sensing object data in multiple adjacent regions. Therefore, step S4 of the present invention can realize object tracking across multiple regions, thereby avoiding the ID switch phenomenon caused by tracking failure when the sensing object data is occluded in a certain region.

[0038] Furthermore, in step S4, to further ensure real-time performance, this invention significantly reduces the amount of data requiring IOU calculation by finding the intersection of the sensing object data sets obtained at different times and filtering out sensing object data with the same ID from multiple sensing object data sets before performing IOU calculation. This better ensures the real-time performance of object tracking. Specific embodiments are described below:

[0039] Please continue reading Figure 2The process involves first finding the intersection of two sets of sensing object data obtained at time T1 and time T2, and then finding the difference between the two sets. This yields the latest extracted sensing object data, IDB3, and the historically cached sensing object data, IDB1. An IOU calculation is then performed on the two sets of sensing object data. If the IOU result is greater than a preset threshold, it indicates that IDB3 and IDB1 are the same sensing object data. IDB3 is then modified to IDB1, thus achieving sensing object data tracking.

[0040] In a further embodiment, when the cached set of perceived object data reaches a preset cache capacity, the first set of perceived object data stored is deleted.

[0041] In a further embodiment, before performing IOU calculation, the present invention can update the location information and heading angle of the sensing object data in the following ways to obtain more accurate location information and heading angle. Specifically, the methods include: determining whether the sensing object data is a connected vehicle based on unique identification information (such as license plate number); acquiring the connected data of the sensing object data when the sensing object data is a connected vehicle, wherein the connected data contains more accurate positioning information and heading angle information; and updating the original location information and heading angle of the sensing object data based on the location information and heading angle in the connected data.

[0042] Figure 4 This is a schematic diagram of a vehicle-road cooperative system according to the present invention. In a second aspect of the present invention, a vehicle-road cooperative system is proposed, such as... Figure 4 As shown, the vehicle-road cooperative system of the present invention includes: a plurality of roadside detectors 100 configured to periodically detect and sense a designated area; a server 200, the server 200 including at least one processor 201 and a memory 202, wherein the memory 202 stores an executable computer program, which, when executed by at least one processor 201, is used to implement the steps of an ID retention method for a vehicle-road cooperative online tracking object as described in any of the above embodiments.

[0043] The above are exemplary embodiments disclosed in this invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments of this invention as defined by the claims. The functions, steps, and / or actions of the methods according to the disclosed embodiments described herein do not need to be performed in any particular order. Furthermore, although the elements disclosed in the embodiments of this invention may be described or claimed individually, they may be understood as multiple unless explicitly limited to a singular number.

[0044] It should be understood that, as used herein, the singular form “a” is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, “and / or” refers to any and all possible combinations of one or more of the associated listed items.

[0045] The embodiment numbers disclosed in the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0046] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples. Within the framework of the invention, technical features of the above embodiments or different embodiments can be combined, and many other variations of different aspects of the invention exist, which are not provided in the details for the sake of brevity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A method for maintaining the ID of a vehicle-road cooperative online tracking object, characterized in that, The method includes: Data on the sensed objects is collected by multiple roadside detectors and stored in corresponding area caches; Retrieve sensor object data from multiple adjacent region caches and reassign a unique ID to each retrieved sensor object data; The acquired sensing object data is deduplicated, including determining whether the sensing object data is located in the detection overlap area based on the location information of the sensing object data; performing position matching on sensing object data in the same detection overlap area; and fusing the ID attributes of the two sensing object data in response to the position deviation of the two sensing object data being less than a preset value, and deleting the sensing object data with the larger ID from the corresponding area cache. The deduplicated sensing object data is fused and stored as a sensing object data set, wherein the sensing object data in the sensing object data set comes from the same moment. Multiple sets of deduplicated sensor object data are cached according to a preset cache capacity. The sensor object data in these multiple sets originates from detection and sensing at different times. The latest and historical sensor object data in these multiple sets are re-divided according to their IDs. Sensor object data with the same ID are assigned to a natural matching set. The remaining sensor object data is divided into a soon-to-disappear set and a newly appearing set based on the order of caching, and IOU data association calculation is performed. The ID of the latest sensor object data is corrected based on the IOU calculation result. Specifically, the soon-to-disappear set is obtained by calculating the difference between the historical sensor object data and the natural matching set, and the newly appearing set is obtained by calculating the difference between the latest sensor object data and the natural matching set.

2. The method for maintaining the ID of a vehicle-road cooperative online tracking object according to claim 1, characterized in that, The roadside detector includes a detection radar and / or a camera, and the sensed object data includes: The circumscribed rectangle, position information, and heading angle.

3. The method for maintaining the ID of a vehicle-road cooperative online tracking object according to claim 1, characterized in that, The process of reassigning a uniquely identifying ID to each acquired sensing object data includes: A unique identifier is reassigned based on the original ID of each of the sensed object data; Initialize the ID attribute, which includes the region where the perceived object data is located, the category to which the perceived object data belongs, and unique identification information; The location information of the sensing object data determines the area where the sensing object data is located, and the bounding rectangle of the sensing object data determines the category to which the sensing object data belongs. The category to which the sensing object data belongs includes designated vehicles and pedestrians, and the unique identification information includes the original ID and / or license plate information of the sensing object data.

4. The method for maintaining the ID of a vehicle-road cooperative online tracking object according to claim 1, characterized in that, The step of performing position matching on the data of sensed objects within the overlapping detection area includes: Location information is matched one by one for the sensing object data belonging to the same category within the overlapping detection area.

5. The method for maintaining the ID of a vehicle-road cooperative online tracking object according to claim 1, characterized in that, The step of correcting the ID of the latest perceived object data based on the IOU calculation result includes: If the IOU calculation result between newly acquired sensory object data in the newly emerging set and historical sensory object data in the soon-to-disappear set is greater than a preset threshold, the ID of the newly acquired sensory object data is corrected to the ID of the corresponding historical sensory object data.

6. The method for maintaining the ID of a vehicle-road cooperative online tracking object according to claim 5, characterized in that, The method further includes: When the cached set of sensory object data reaches the preset cache capacity, the first set of sensory object data stored is deleted.

7. A method for maintaining the ID of a vehicle-road cooperative online tracking object according to claim 5 or 6, characterized in that, The IOU calculation includes: IOU calculation is performed based on the location information, heading angle, and overlapping area of ​​the circumscribed rectangle of the newly acquired sensing object data and historical sensing object data.

8. The method for maintaining the ID of a vehicle-road cooperative online tracking object according to claim 7, characterized in that, The method also includes performing data association calculations beforehand. Based on the unique identification information, determine whether the data of the sensed object is a connected vehicle; When the perceived object data is a connected vehicle, the connected data of the perceived object data is acquired. The original location information and heading angle of the sensed object data are updated based on the location information and heading angle in the network data.

9. A vehicle-road cooperative system, characterized in that, include: Multiple roadside detectors are configured to periodically detect and sense designated areas; The server includes at least one processor and a memory storing an executable computer program, which, when executed by the at least one processor, is used to implement the steps of an ID retention method for a vehicle-road cooperative online tracking object as described in any one of claims 1-8.

Images