A cross-region target tracking method based on trajectory point matching

Abstract

The application discloses a cross-region target tracking method based on trajectory point matching, which obtains target state information of a target under different radar and visual conditions by analyzing radar and visual overlapping perception region target structured data; carries out trajectory feature point extraction on the target state information under different radar and visual conditions of the overlapping perception region, carries out registration by using the mutual relationship between the feature points, and determines an upstream and downstream radar and visual perception target position conversion function; calculates the distance between a pre-matching target trajectory point of downstream radar and visual perception and a to-be-matched target trajectory point of upstream radar and visual perception, determines the corresponding relationship between the upstream and downstream radar and visual perception targets in the overlapping perception region according to a maximum matching probability method, and completes cross-perception region target tracking. The application increases the completeness of target perception information, cooperates different radar and visual conditions for continuous traffic perception, and provides a new method for intelligent traffic management.

Description

Technical Field

[0001] This invention belongs to the field of intelligent traffic detection technology, and in particular to a cross-regional target tracking method based on trajectory point matching, especially a target tracking method based on multiple radar sights having a common detection area. Background Technology

[0002] In recent years, with the acceleration of the intelligentization of urban transportation, there is a greater demand for high-grade roads that can provide richer and more comprehensive perception information. As a roadside perception device that is less affected by the environment and has a longer detection range, radar vision has been widely used in the construction of high-grade roads. At the same time, radar vision combines the advantages of millimeter-wave radar and video, and can obtain high-precision detection results for both static and dynamic information of objects.

[0003] During the construction of high-grade roads, the deployment of radar-based surveillance systems (Radar cameras) provides the traffic system with abundant roadside data. When detecting vehicles, the radar cameras use random numbers to identify them; this identification is unique only within a certain timeframe. The identification of the same target is unrelated across different radar cameras. Therefore, during continuous road detection using radar cameras, the problem arises where the same target cannot be continuously tracked across different radar camera sensing areas. Trajectory deduplication and tracking are crucial aspects of constructing high-grade roads.

[0004] Utilizing the trajectory features of targets in overlapping sensing areas is a crucial method for continuous target tracking across sensing areas. In high-grade road deployments, radar detectors (Radar cameras) are often deployed continuously, with overlapping sensing areas between upstream and downstream radar camera detection zones. At these overlapping zones, both upstream and downstream radar cameras can detect the same target. By using the trajectory features of the detected target for target matching, continuous dynamic motion trajectories of the target across sensing areas can be obtained, thus improving the perception capabilities of high-grade roads. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies where different radar cameras give different identifiers to the same target. It proposes a cross-regional target tracking method based on trajectory point matching. This method uses overlapping sensing area data from upstream and downstream radar cameras, dynamic motion feature data of targets within the overlapping sensing area, and uses feature point registration and trajectory point matching to match the targets sensed by upstream and downstream radar cameras, thus forming cross-regional target sensing and tracking.

[0006] The technical solution for achieving the objective of this invention is: a cross-regional target tracking method based on trajectory point matching, the method comprising the following steps:

[0007] Step 1: Collect and parse the target structured data output by the radar vision system to obtain the position, velocity, heading angle, time, and local information of the perceived target under the current radar vision system.

[0008] Step 2: Compare the current boundary of the radar vision sensing area with the location of the sensed target, and dynamically update the boundary of the radar vision sensing area based on the relationship between the target location and the boundary of the sensing area.

[0009] Step 3: Compare the boundaries of the upstream and downstream radar vision sensing areas, and obtain the overlapping sensing area between the upstream and downstream radar vision areas based on the relative positions of the sensing area boundaries.

[0010] Step 4: Compare the target location with the boundary of the overlapping sensing area. For targets outside the overlapping sensing area, if the globally unique identifier is empty, generate a unique identifier based on the target detection time and radar number. For targets inside the overlapping sensing area, construct the upstream and downstream target trajectory datasets and output the targets with unique identifiers.

[0011] Step 5: Analyze the target datasets of upstream and downstream radar vision perception in the overlapping perception area, find the multi-time feature point set of downstream radar vision perception targets until the feature point set meets the registration requirements, and dynamically update the corresponding point set of upstream radar vision perception targets waiting for registration based on the feature point set selected by the downstream radar vision perception targets.

[0012] Step 6: Register the feature point set of downstream radar sight at multiple time points in the overlapping sensing area with the corresponding point set selected by upstream radar sight to establish an upstream and downstream trajectory point registration and transformation matrix.

[0013] Step 7: Filter downstream radar vision pre-matching targets. When the number of historical trajectory points of the collected targets reaches a certain threshold, select the historical trajectory point set of the pre-matching target, establish a target matcher, and calculate the degree of similarity between all selected historical trajectory points and the upstream radar vision collected target to be matched.

[0014] Step 8: Filter upstream and downstream sensing target matchers, complete the matching of upstream and downstream radar sensing targets, the downstream radar pre-matched target obtains the unique identifier of the corresponding upstream radar target to be matched after successful matching, the upstream radar ends the output of the matched target, and the downstream radar begins to output the matched target data.

[0015] Furthermore, the specific process of step 2 includes:

[0016] Construct a boundary table for all radar-sensing areas according to the lane direction. The boundary is the latitude and longitude of the nearest and farthest points in the corresponding lane direction. The initial boundary is initialized according to the installation angle of the radar.

[0017] A perpendicular line is drawn from the boundary trajectory point of the sensing area, and the road width corresponding to the movement of the boundary trajectory point along the perpendicular line is used as the coordinates of the boundary discrimination point of the sensing area.

[0018] Based on the dynamic target status information of the radar vision sensing, the boundary of the radar vision sensing area is updated in real time.

[0019] Furthermore, the formula for calculating the boundary of the sensing area is:

[0020]

[0021] In the formula, x boundary and y boundary The boundary of the radar vision sensing area; x o and y o The location of the target trajectory point collected at the boundary; n represents the number of lanes from the trajectory point to the road edge; α is the conversion coefficient between road width and longitude; β is the conversion coefficient between road width and latitude; width represents the lane width.

[0022] Furthermore, step 3 updates the overlapping sensing area between upstream and downstream radar sensors. The specific process includes:

[0023] Based on the planned installation location and detection direction of the radar screen, a preliminary judgment is made on all possible upstream and downstream relationships between different radar screens, and a radar screen upstream and downstream relationship table is established.

[0024] Compare the boundaries of radar vision systems that may have upstream and downstream relationships at fixed intervals to determine whether there is an overlapping sensing area between two radar vision systems.

[0025] For radar-based cameras with overlapping sensing areas, a target matching notification device is established. For any target entering the radar-based camera's sensing area, target tracking across overlapping sensing areas is initiated.

[0026] Furthermore, in step 4, the specific processing procedure for targets within the overlapping sensing area includes:

[0027] Construct upstream and downstream target trajectory datasets. The target trajectory information includes longitude, latitude, speed, heading angle, and acquisition time.

[0028] To determine the target location, for targets acquired by upstream radar, if they are not located in the overlapping sensing area and do not have a unique identifier, a unique identifier for the target is generated according to the globally unique ID generation algorithm and the target is output; if they are not located in the overlapping sensing area but have a unique identifier, the target is output directly; if they are located in the overlapping sensing area, the target is output and the target parameters are input into the upstream target trajectory dataset and the upstream target matcher.

[0029] If the target does not have a globally unique ID, the target will not be output until a match is found.

[0030] Furthermore, to avoid errors in the radar vision sensing system, target position transformation is performed; trajectory feature points in the downstream radar vision sensing area of ​​the overlapping sensing area are selected, and the corresponding trajectory feature points at the upstream radar vision location are determined to construct a target position transformation matrix.

[0031] Furthermore, step 5, which involves finding registration feature points in the overlapping sensing region, specifically includes the following process:

[0032] Establish a feature point matcher to store the feature points selected by upstream and downstream radars, and select feature points based on the trajectory dataset stored by downstream radars;

[0033] Trajectory datasets stored in downstream radar vision are traversed, and several targets are selected from the leftmost and rightmost lanes respectively. These targets are used as registration points to obtain the trajectory lines of these targets.

[0034] Based on the occurrence time and location of all trajectory points within the trajectory line, select the same number of the leftmost lane targets and the rightmost lane targets from the target trajectory point set collected by upstream radar.

[0035] Further, step 6, which involves registering the upstream and downstream radar sensing regions based on the registration feature points, specifically includes the following process:

[0036] Establish the registration point set and corresponding point set for the leftmost and rightmost sides of the road in the upstream and downstream radar vision sensing areas;

[0037] The error exists when upstream and downstream radar cameras detect the same location area. An objective function is established based on minimizing the distance between selected trajectory lines near this area, resulting in the registration transformation function for the sensing location from downstream to upstream radar cameras, represented by the following registration transformation matrix:

[0038]

[0039] In the formula, lon a and lat a For the converted longitude and latitude; lon a-1 and lat a-1 For the longitude and latitude before conversion, θ 11 θ 12 θ 21 and θ 22 b1 and b2 are the rotation matrix parameters for latitude and longitude transformation; b1 and b2 are the supplementary terms for the lateral and vertical errors after latitude and longitude transformation.

[0040] Furthermore, step 7, which involves matching targets acquired by upstream and downstream radar cameras within the overlapping sensing area, specifically includes the following process:

[0041] Filtering pre-matched targets: If a target is captured by downstream radar more than a set threshold in the overlapping sensing area and the target is moving in the direction of the road, then the target is stored in the pre-matched target set.

[0042] For the pre-matched target collected by the downstream radar, all trajectory points of the target within a certain time threshold are traversed, and the registration transformation matrix is ​​used to calculate the target position under the perception of the upstream radar.

[0043] Determine fixed thresholds for the three dimensions of the target's location after transformation: the nearest time, longitude, and latitude, and establish the maximum matching space;

[0044] Select the trajectory points of all targets to be matched within the upstream radar field of the maximum matching space, and calculate the average distance in time, longitude, and latitude between the trajectory points of all targets to be matched and the pre-matched targets within the matching space;

[0045]

[0046]

[0047]

[0048] In the formula, Time distance,k Longitude distance,k Latitude distance,k , respectively, represent the average time distance, average longitude distance, and average latitude distance between the downstream radar-sensing pre-matching target and the upstream radar-sensing target to be matched k within the fixed matching distance threshold space; n is the number of targets to be matched within the fixed matching distance threshold space; i is the i-th trajectory point among the k-th target to be matched; Time i Longitude i Latitude i Let Time be the perceived time, longitude, and latitude of the i-th trajectory point of the target k to be matched. target Longitude target Latitude target The perceived time, longitude, and latitude of the pre-matched target for downstream sensing;

[0049] Based on road alignment and radar delay, the weights of time, longitude, and latitude are determined. The average distance between all trajectory points of the target to be matched and the pre-matched target within the matching space is calculated. The target to be matched with the smallest average distance is taken as the corresponding matching target of the pre-matched target. The comprehensive distance between the pre-matched target and the target to be matched is given by the following equation:

[0050]

[0051] In the formula, Index distance,k Let k be the combined distance between the target to be matched (k) and the pre-matched target; k1, k2, and k3 are the weights of time, longitude, and latitude, respectively, and k1 + k2 + k3 = 1.

[0052] Furthermore, step 8, which describes the continuous tracking of upstream and downstream targets, specifically includes the following process:

[0053] Calculate the probability of each target to be matched corresponding to the downstream perceived pre-matching target; the probability of each target to be matched being successful is:

[0054]

[0055] In the formula, P k n represents the probability that the upstream target k to be matched is successfully matched; k This represents the number of times the upstream target k was successfully matched, and N represents the total number of matches.

[0056] Determine if the probability of the target being matched is greater than a set threshold; if so, complete the matching.

[0057] After the upstream and downstream targets are matched, the upstream radar stops outputting data to the matched targets, and the downstream radar starts outputting data to the matched targets.

[0058] Compared with the prior art, the significant advantages of this invention are:

[0059] 1) Based on the collected target information, the present invention dynamically updates the upstream and downstream relationships and overlapping sensing area positions between different radar detectors, which can avoid the influence of the environment such as rain and fog on the range of radar detector sensing area.

[0060] 2) This invention establishes upstream and downstream radar-visual relationships using a one-way road segment approach, which avoids the problem of different perception boundaries between incoming and outgoing directions for radar-visual systems.

[0061] 3) Before matching upstream and downstream radar targets, this invention uses feature points to register the trajectories of upstream and downstream targets. This can avoid the problem of increased target position error when both upstream and downstream radars have certain perception errors, as well as the problem of deviation in the calibrated road perception area due to ground shaking and wind.

[0062] 4) This invention utilizes the target trajectory features of overlapping sensing areas to perform upstream and downstream radar-sensing target matching, enabling cross-regional tracking of targets in different sensing areas, increasing the integrity of target sensing information, and coordinating different radar-sensing systems for continuous traffic sensing, providing a new method for intelligent traffic management.

[0063] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating a cross-regional target tracking method based on trajectory point matching, provided as an embodiment.

[0065] Figure 2 A schematic diagram of the boundary of the overlapping area between upstream and downstream radar lines provided in one embodiment.

[0066] Figure 3 A schematic diagram of the registration of the upstream and downstream overlapping areas of the radar provided in one embodiment. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0068] It should be noted that if the embodiments of the present invention involve descriptions such as "first" and "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include at least one of those features. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0069] In one embodiment, a cross-regional target tracking method based on trajectory point matching is provided. This method collects radar target data, identifies overlapping radar sensing areas, and utilizes the trajectory point features of upstream and downstream radar sensing targets within the sensing areas to achieve target matching between upstream and downstream radar sensing areas. Figure 1 As shown. It should be noted that before completing target matching, the upstream and downstream relationships of the radar sights, and the overlapping area of ​​the upstream and downstream radar sights, need to be dynamically updated to reduce the error in the sensing area registration. The specific steps are as follows:

[0070] Step 1: Collect and parse the target structured data output by the radar vision system. The target information includes the target's perceived position, speed, heading angle, time, and local identifiers under the current radar vision system.

[0071] Step 2: Determine the boundary of the radar vision sensing area;

[0072] The boundaries of the radar-sensing area are dynamically updated based on trajectory data. The specific steps are as follows:

[0073] Step 2-1: Construct a boundary table for all radar-sensing areas according to the lane direction. The boundaries are the latitude and longitude of the nearest and farthest points. The initial boundaries are initialized according to the radar installation angle.

[0074] Step 2-2: Draw a perpendicular line from the boundary trajectory point of the sensing area. Move the boundary trajectory point along the perpendicular line to the corresponding road width to obtain the coordinates of the sensing area boundary discrimination point. The calculation process of the sensing area boundary is as follows:

[0075]

[0076] In the formula, x boundary and y boundary The boundary of the radar vision sensing area; x o and y o The location of the target trajectory point collected at the boundary; n represents the number of lanes from the trajectory point to the road edge; α is the conversion coefficient between road width in meters and longitude; β is the conversion coefficient between road width in meters and latitude; width represents the lane width;

[0077] Steps 2-3: Update the boundary of the radar-sensing area based on the dynamic radar-sensing target status information. The boundary of the radar-sensing area is dynamically updated according to the target location and the boundary of the sensing area.

[0078] Step 3: Identify the overlapping sensing areas between upstream and downstream radar cameras;

[0079] Figure 2 The status of overlapping radar sensing areas on a one-way, two-lane high-grade road section is given, and the specific process of identifying overlapping sensing areas is as follows:

[0080] Step 3-1: Based on the predetermined installation location and detection direction of the radar screen, preliminarily determine all possible upstream and downstream relationships between different radar screens and establish a radar screen upstream and downstream relationship table;

[0081] Step 3-2: At fixed intervals, compare the boundary areas of the radar cameras that may have upstream and downstream relationships to determine whether there is an overlapping sensing area between the two radar cameras.

[0082] Step 3-3: For two radar cameras with overlapping areas, establish a target matching notification device to enable cross-sensing area matching for any target entering the sensing area of ​​the upstream and downstream radar cameras.

[0083] Step 4: Output the target in the overlapping sensing area to be matched;

[0084] Step 4-1: For upstream and downstream radar cameras with overlapping sensing areas, construct upstream and downstream target trajectory datasets and store target trajectory data information in the form of a hash table. The key is the ID of the target under the corresponding radar camera, and the value is the motion state parameters collected by the target, including longitude, latitude, speed, heading angle, collection time, and target global ID.

[0085] Step 4-2, Target Location Judgment: For targets collected by upstream radar, if they are not located in the overlapping area and have no unique identifier, generate a unique identifier for the target based on the globally unique ID generation algorithm and output the target; if they are not located in the overlapping area and have a unique identifier, output the target directly; if they are located in the overlapping area, output the target and store the target parameters in the upstream target dataset to be matched.

[0086] Step 4-3: For downstream targets that have upstream targets, if the targets are located in overlapping areas and have no unique identifiers, store the target parameters in the downstream pre-matching target set. Do not output the target if the matching is not completed.

[0087] Step 4-4: When the upstream and downstream radar targets have not completed target matching, only the target collected by the upstream radar is output, and the target that the downstream radar has not successfully matched is not output; after the upstream and downstream radar sensing targets are matched, the target collected by the upstream radar is not output, and the target that the downstream radar has successfully matched is output.

[0088] Step 5: Complete the lookup of registration feature points in the overlapping sensing region;

[0089] Step 5-1: Establish a feature point matcher to store the feature points selected by upstream and downstream radar cameras, and select feature points starting from the trajectory dataset stored by downstream radar cameras.

[0090] Step 5-2: Traverse the trajectory dataset stored in the downstream radar vision system, select a certain number of targets in the leftmost and rightmost test lanes respectively, and use these targets as registration points to obtain the trajectory lines of these targets.

[0091] Step 5-3: Based on the occurrence time and location of all trajectory points within the trajectory line, select the closest leftmost lane target and rightmost lane target of the same number from the target trajectory dataset collected by upstream radar vision.

[0092] Step 6: Register the upstream and downstream radar sensing areas of the overlapping sensing area;

[0093] Step 6-1: Establish the registration point set and corresponding point set for the leftmost and rightmost sides of the road in the upstream and downstream radar vision sensing areas;

[0094] Step 6-2: Calculate the error when two radar cameras monitor the same location area. Establish the objective function based on the shortest distance between the selected trajectory lines near the area to obtain the registration transformation function for the sensing position from the downstream radar camera to the upstream radar camera. The transformation calculation method is given by the following equation:

[0095]

[0096] In the formula, lon a and lat aFor the converted longitude and latitude; lon a-1 and lat a-1 For the longitude and latitude before conversion, θ 11 θ 12 θ 21 and θ 22 b1 and b2 are the rotation matrix parameters for latitude and longitude transformation; b1 and b2 are the supplementary terms for horizontal and vertical errors after latitude and longitude transformation.

[0097] Figure 3 The process of registration between upstream and downstream radar sensors within the overlapping sensing area is presented.

[0098] Step 7: Match the targets acquired by upstream and downstream radar cameras within the overlapping area. The specific process includes:

[0099] Step 7-1, Screening pre-matched targets: If the number of times a target is captured by downstream radar in the overlapping perception area exceeds a certain threshold, and the target's motion state is moving in the direction of the road, false identification is excluded.

[0100] Step 7-2: For the selected downstream target collected by radar, traverse all the target trajectory points within a certain time threshold, use the trajectory points as matching points, and use the registration transformation matrix to calculate the possible positions under upstream radar sensing.

[0101] Step 7-3: Establish fixed thresholds for the three dimensions of the target location: the nearest time, longitude, and latitude, and create the maximum matching space;

[0102] Step 7-4: Select all target trajectory points collected by upstream radar within the maximum matching space, and calculate the average distance between all targets and the pre-matching target in terms of time, longitude, and latitude within the matching space. The specific calculation equation is as follows:

[0103]

[0104]

[0105]

[0106] In the formula, Time distance,k Longitude distance,k Latitude distance,k , respectively, represent the average time distance, average longitude distance, and average latitude distance between the downstream radar-sensing pre-matched target trajectory point and the upstream radar-sensing target k to be matched within the fixed matching distance threshold space; n is the number of targets to be matched within the fixed matching distance threshold space; i is the i-th trajectory point among the k-th target to be matched; Time i Longitude i Latitudei Let Time be the perceived time, longitude, and latitude of the i-th trajectory point of the target k to be matched. target Longitude target Latitude target The perceived time, longitude, and latitude of the pre-matched target for downstream sensing;

[0107] Step 7-5: Based on the road alignment and radar delay, determine the weights of time, longitude, and latitude, calculate the average distance between all pre-matched targets and the target to be matched within the matching space, and use the target to be matched corresponding to the minimum average distance as the completed matching target of the pre-matched target. The comprehensive distance between the downstream radar-sensed pre-matched target and the upstream radar-sensed target to be matched is given by the following equation:

[0108]

[0109] In the formula, Index distance,k The total distance between the pre-matched target and the target to be matched k is given by k1, k2, and k3, which are the weights of time, longitude, and latitude, respectively, and k1+k2+k3=1.

[0110] Step 8: Complete continuous tracking of upstream and downstream radar targets;

[0111] Step 8-1: Calculate the probability of each target to be matched corresponding to the downstream pre-matched target; the probability of each upstream target to be matched being successfully matched is given by the following equation:

[0112]

[0113] In the formula, P k n represents the probability that the upstream target k to be matched is successfully matched; k This represents the number of times the upstream target k was successfully matched, and N represents the total number of matches.

[0114] Step 8-2: Once the probability of the target being successfully matched is greater than the set threshold, the matching is completed.

[0115] Step 8-3: After the upstream and downstream radar cameras complete the matching of their corresponding targets, the upstream radar camera stops outputting data to the matched targets, and the downstream radar camera starts outputting data to the matched targets.

[0116] In one embodiment, a cross-regional target tracking system based on trajectory point matching is provided, the system comprising:

[0117] The first module is used to collect and parse the target structured data output by the radar vision system to obtain the position, speed, heading angle, time, and identification of the perceived target.

[0118] The second module is used to compare the current boundary of the radar vision sensing area with the location of the sensed target, and dynamically update the boundary of the radar vision sensing area based on the relationship between the target location and the boundary of the sensing area.

[0119] The third module is used to compare the boundaries of the upstream and downstream radar vision sensing areas, and to obtain and update the overlapping sensing areas between the upstream and downstream radar vision areas in real time based on the relative positions of the sensing area boundaries.

[0120] The fourth module is used to output the targets in the overlapping sensing area to be matched: compare the target position with the boundary of the overlapping sensing area, form a globally unique identifier for targets outside the overlapping sensing area, construct the upstream and downstream target trajectory dataset for targets inside the overlapping sensing area, and output the targets with unique identifiers.

[0121] The fifth module is used to find registration feature points in the overlapping sensing region;

[0122] The sixth module is used to register the positions of the upstream and downstream radar vision sensing areas of the overlapping sensing area based on the registration feature points.

[0123] The seventh module is used to match targets collected by upstream and downstream radar in the overlapping sensing area: filter downstream radar sensing pre-matching targets, and when the number of historical trajectory points of the collected targets reaches a preset threshold, select the set of historical trajectory points of the pre-matching target, establish a target matcher, and calculate the proximity of all selected historical trajectory points to the upstream radar-collected target to be matched.

[0124] The eighth module is used to realize continuous tracking of upstream and downstream radar targets: filter upstream and downstream sensing target matchers, complete the matching of upstream and downstream radar sensing targets, the downstream radar pre-matched target obtains the unique identifier of the corresponding upstream radar target to be matched after successful matching, the upstream radar ends the output of the matched target, and the downstream radar begins to output the target data of the matched target.

[0125] Specific limitations regarding the cross-regional target tracking system based on trajectory point matching can be found in the limitations of the cross-regional target tracking method based on trajectory point matching above, and will not be repeated here. Each module in the aforementioned cross-regional target tracking system based on trajectory point matching can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0126] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0127] Step 1: Collect and parse the target structured data output by the radar vision system to obtain the position, velocity, heading angle, time, and identifier of the perceived target;

[0128] Step 2: Compare the current boundary of the radar vision sensing area with the location of the sensed target, and dynamically update the boundary of the radar vision sensing area based on the relationship between the target location and the boundary of the sensing area.

[0129] Step 3: Compare the boundaries of the upstream and downstream radar vision sensing areas, and obtain and update the overlapping sensing areas between the upstream and downstream radar vision areas in real time based on the relative positions of the sensing area boundaries.

[0130] Step 4, output the target in the overlapping sensing area to be matched: compare the target position with the boundary of the overlapping sensing area, form a globally unique identifier for targets outside the overlapping sensing area, construct the upstream and downstream target trajectory dataset for targets inside the overlapping sensing area, and output the targets with unique identifiers.

[0131] Step 5: Locate registration feature points in the overlapping sensing region;

[0132] Step 6: Based on the registration feature points, register the positions of the upstream and downstream radar vision sensing areas of the overlapping sensing area;

[0133] Step 7: Match the targets collected by upstream and downstream radar in the overlapping sensing area: Filter the downstream radar sensing pre-matching targets. When the number of historical trajectory points of the collected targets reaches a preset threshold, select the set of historical trajectory points of the pre-matching targets, establish a target matcher, and calculate the proximity of all selected historical trajectory points to the upstream radar-collected target to be matched.

[0134] Step 8, achieve continuous tracking of upstream and downstream radar targets: filter upstream and downstream sensing target matchers, complete the matching of upstream and downstream radar sensing targets, the downstream radar pre-matched target obtains the unique identifier of the corresponding upstream radar target to be matched after successful matching, the upstream radar ends the output of the matched target, and the downstream radar begins to output the matched target data.

[0135] For specific limitations on each step, please refer to the limitations of the cross-regional target tracking method based on trajectory point matching mentioned above, which will not be repeated here.

[0136] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0137] Step 1: Collect and parse the target structured data output by the radar vision system to obtain the position, velocity, heading angle, time, and identifier of the perceived target;

[0138] Step 2: Compare the current boundary of the radar vision sensing area with the location of the sensed target, and dynamically update the boundary of the radar vision sensing area based on the relationship between the target location and the boundary of the sensing area.

[0139] Step 3: Compare the boundaries of the upstream and downstream radar vision sensing areas, and obtain and update the overlapping sensing areas between the upstream and downstream radar vision areas in real time based on the relative positions of the sensing area boundaries.

[0140] Step 4, output the target in the overlapping sensing area to be matched: compare the target position with the boundary of the overlapping sensing area, form a globally unique identifier for targets outside the overlapping sensing area, construct the upstream and downstream target trajectory dataset for targets inside the overlapping sensing area, and output the targets with unique identifiers.

[0141] Step 5: Locate registration feature points in the overlapping sensing region;

[0142] Step 6: Based on the registration feature points, register the positions of the upstream and downstream radar vision sensing areas of the overlapping sensing area;

[0143] Step 7: Match the targets collected by upstream and downstream radar in the overlapping sensing area: Filter the downstream radar sensing pre-matching targets. When the number of historical trajectory points of the collected targets reaches a preset threshold, select the set of historical trajectory points of the pre-matching targets, establish a target matcher, and calculate the proximity of all selected historical trajectory points to the upstream radar-collected target to be matched.

[0144] Step 8, achieve continuous tracking of upstream and downstream radar targets: filter upstream and downstream sensing target matchers, complete the matching of upstream and downstream radar sensing targets, the downstream radar pre-matched target obtains the unique identifier of the corresponding upstream radar target to be matched after successful matching, the upstream radar ends the output of the matched target, and the downstream radar begins to output the matched target data.

[0145] For specific limitations on each step, please refer to the limitations of the cross-regional target tracking method based on trajectory point matching mentioned above, which will not be repeated here.

[0146] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention without departing from its spirit and scope should be included within the protection scope of the present invention.

Claims

1. A cross-regional target tracking method based on trajectory point matching, characterized in that, The method includes the following steps: Step 1: Collect and parse the target structured data output by the radar vision system to obtain the position, velocity, heading angle, time, and local identifiers of the perceived target under the current radar vision system. Step 2: Compare the current boundary of the radar vision sensing area with the location of the sensed target, and dynamically update the boundary of the radar vision sensing area based on the relationship between the target location and the boundary of the sensing area. Step 3: Compare the boundaries of the upstream and downstream radar vision sensing areas, and obtain and update the overlapping sensing areas between the upstream and downstream radar vision areas in real time based on the relative positions of the sensing area boundaries. Step 4, output the target in the overlapping sensing area to be matched: compare the target position with the boundary of the overlapping sensing area. For targets outside the overlapping sensing area, generate a unique identifier based on the target detection time and radar number when the global unique identifier is empty. For targets within the overlapping sensing area, construct the upstream and downstream target trajectory datasets and output the targets with unique identifiers. Step 5: Locate registration feature points in the overlapping sensing region; Step 6: Based on the registration feature points, register the positions of the upstream and downstream radar vision sensing areas of the overlapping sensing area; Step 7: Match the targets collected by upstream and downstream radar in the overlapping sensing area: Filter the downstream radar sensing pre-matching targets. When the number of historical trajectory points of the collected targets reaches a preset threshold, select the set of historical trajectory points of the pre-matching targets, establish a target matcher, and calculate the proximity of all selected historical trajectory points to the upstream radar-collected target to be matched. Step 8, achieve continuous tracking of upstream and downstream radar targets: filter upstream and downstream sensing target matchers, complete the matching of upstream and downstream radar sensing targets, the downstream radar pre-matched target obtains the unique identifier of the corresponding upstream radar target to be matched after successful matching, the upstream radar ends the output of the matched target, and the downstream radar begins to output the matched target data. Step 5, which involves finding registration feature points in the overlapping sensing region, specifically includes the following process: Establish a feature point matcher to store the feature points selected by upstream and downstream radars, and select feature points based on the trajectory dataset stored by downstream radars; Trajectory datasets stored in downstream radar vision are traversed, and several targets are selected from the leftmost and rightmost lanes respectively. These targets are used as registration points to obtain the trajectory lines of these targets. Based on the occurrence time and location of all trajectory points within the trajectory line, select the same number of the leftmost lane targets and the rightmost lane targets from the target trajectory point set collected by upstream radar.

2. The cross-regional target tracking method based on trajectory point matching according to claim 1, characterized in that, Step 2 includes the following specific steps: Construct a boundary table for all radar-sensing areas according to the lane direction. The boundary is the latitude and longitude of the nearest and farthest points in the corresponding lane direction. The initial boundary is initialized according to the installation angle of the radar. A perpendicular line is drawn from the boundary trajectory point of the sensing area, and the road width corresponding to the movement of the boundary trajectory point along the perpendicular line is used as the coordinates of the boundary discrimination point of the sensing area. Based on the dynamic target status information of the radar vision sensing, the boundary of the radar vision sensing area is updated in real time.

3. The cross-regional target tracking method based on trajectory point matching according to claim 2, characterized in that, The formula for calculating the boundary of the sensing area is: In the formula, x boundary and y boundary The boundary of the radar vision sensing area; x o and y o The location of the target trajectory point collected at the boundary; n represents the number of lanes from the trajectory point to the road edge; α is the conversion coefficient between road width and longitude; β is the conversion coefficient between road width and latitude; width represents the lane width.

4. The cross-regional target tracking method based on trajectory point matching according to claim 1, characterized in that, Step 3 involves updating the overlapping sensing area between upstream and downstream radar sensors. The specific process includes: Based on the planned installation location and detection direction of the radar screen, a preliminary judgment is made on all possible upstream and downstream relationships between different radar screens, and a radar screen upstream and downstream relationship table is established. Compare the boundaries of radar vision systems that may have upstream and downstream relationships at fixed intervals to determine whether there is an overlapping sensing area between two radar vision systems. For radar-based cameras with overlapping sensing areas, a target matching notification device is established. For any target entering the radar-based camera's sensing area, target tracking across overlapping sensing areas is initiated.

5. The cross-regional target tracking method based on trajectory point matching according to claim 4, characterized in that, Step 4 involves the following specific processing steps for targets within the overlapping sensing area: Construct upstream and downstream target trajectory datasets. The target trajectory information includes longitude, latitude, speed, heading angle, and acquisition time. To determine the target location, for targets acquired by upstream radar, if they are not located in the overlapping sensing area and do not have a unique identifier, a unique identifier for the target is generated according to the globally unique ID generation algorithm and the target is output; if they are not located in the overlapping sensing area but have a unique identifier, the target is output directly; if they are located in the overlapping sensing area, the target is output and the target parameters are input into the upstream target trajectory dataset and the upstream target matcher. If the target does not have a globally unique ID, the target will not be output until a match is found.

6. The cross-regional target tracking method based on trajectory point matching according to claim 5, characterized in that, The target trajectory information is stored in the form of a hash table, where the key is the target's ID under the corresponding radar, and the value is the target trajectory information, including longitude, latitude, speed, heading angle, acquisition time, and target global ID.

7. The cross-regional target tracking method based on trajectory point matching according to claim 6, characterized in that, Step 6 involves registering the upstream and downstream radar sensing regions of the overlapping sensing region based on the registration feature points. The specific process includes: Establish the registration point set and corresponding point set for the leftmost and rightmost sides of the road in the upstream and downstream radar vision sensing areas; The error exists when upstream and downstream radar cameras detect the same location area. An objective function is established based on minimizing the distance between selected trajectory lines near this area, resulting in the registration transformation function for the sensing location from downstream to upstream radar cameras, represented by the following registration transformation matrix: In the formula, lon a and lat a For the converted longitude and latitude; lon a-1 and lat a-1 For the longitude and latitude before conversion, θ 11 θ 12 θ 21 and θ 22 b1 and b2 are the rotation matrix parameters for latitude and longitude transformation; b1 and b2 are the supplementary terms for the lateral and vertical errors after latitude and longitude transformation.

8. The cross-regional target tracking method based on trajectory point matching according to claim 7, characterized in that, Step 7, which involves matching targets acquired by upstream and downstream radar cameras within the overlapping sensing area, specifically includes the following process: Filtering pre-matched targets: If a target is captured by downstream radar more than a set threshold in the overlapping sensing area and the target is moving in the direction of the road, then the target is stored in the pre-matched target set. For the pre-matched target collected by the downstream radar, all trajectory points of the target within a certain time threshold are traversed, and the registration transformation matrix is ​​used to calculate the target position under the perception of the upstream radar. Determine fixed thresholds for the three dimensions of the target's location after transformation: the nearest time, longitude, and latitude, and establish the maximum matching space; Select the trajectory points of all targets to be matched within the upstream radar field of the maximum matching space, and calculate the average distance in time, longitude, and latitude between the trajectory points of all targets to be matched and the pre-matched targets within the matching space; In the formula, Time distance,k Longitude distance,k Latitude distance,k , respectively, represent the average time distance, average longitude distance, and average latitude distance between the downstream radar-sensing pre-matching target and the upstream radar-sensing target to be matched k within the fixed matching distance threshold space; n is the number of targets to be matched within the fixed matching distance threshold space; i is the i-th trajectory point among the k-th target to be matched; Time i Longitude i Latitude i Let Time be the perceived time, longitude, and latitude of the i-th trajectory point of the target k to be matched. target Longitude target Latitude target The perceived time, longitude, and latitude of the pre-matched target for downstream sensing; Based on road alignment and radar delay, the weights of time, longitude, and latitude are determined. The average distance between all trajectory points of the target to be matched and the pre-matched target within the matching space is calculated. The target to be matched with the smallest average distance is taken as the corresponding matching target of the pre-matched target. The comprehensive distance between the pre-matched target and the target to be matched is given by the following equation: In the formula, Index distance,k Let k be the combined distance between the target to be matched k and the pre-matched target; k1, k2, and k3 are the weights of time, longitude, and latitude, respectively, and k1+k2+k3=1.

9. The cross-regional target tracking method based on trajectory point matching according to claim 8, characterized in that, Step 8 describes the implementation of continuous tracking of upstream and downstream targets using radar, and the specific process includes: Calculate the probability of each target to be matched corresponding to the downstream perceived pre-matching target; the probability of each target to be matched being successful is: In the formula, P k n represents the probability that the upstream target k to be matched is successfully matched; k This represents the number of times the upstream target k was successfully matched, and N represents the total number of matches. Determine if the probability of the target being matched is greater than a set threshold; if so, complete the matching. After the upstream and downstream targets are matched, the upstream radar stops outputting data to the matched targets, and the downstream radar starts outputting data to the matched targets.

Images