A radar and visual fusion method based on static target correction

Abstract

The application discloses a radar and video fusion method based on static target correction. The method comprises the following steps: acquiring a millimeter wave radar target tracking result and a video target detection result; projecting the millimeter wave radar target tracking result to a video picture coordinate system, and combining a target tracking trajectory to judge a target motion state; the target motion state is divided into four categories: motion, static, quasi-static and disappearance; correcting the targets in the static and quasi-static states; re-matching the newly-recognized targets of the millimeter wave radar, removing repeated targets, and adding newly-recognized targets; and updating the millimeter wave radar target tracking result. The application can effectively identify and correct the static and quasi-static targets that cannot be detected by the millimeter wave radar.

Description

Technical Field

[0001] This invention belongs to the fields of intelligent transportation and target perception, and in particular, it is a radar-visual fusion method based on static target correction. Background Technology

[0002] Rayvision-Ray fusion perception technology is a new technology that combines millimeter-wave radar and video to achieve target perception. This new technology fully utilizes the anti-environmental interference capability of millimeter-wave radar to achieve all-weather perception. At the same time, by leveraging the visualization and structuring capabilities of video, it can extract target attributes and license plate information, greatly enriching the dimensions of millimeter-wave radar perception data.

[0003] Traditional millimeter-wave radar lacks the ability to extract elevation information and has poor low-speed target identification capabilities, failing to effectively distinguish between stationary vehicles and roadside signs. A common approach is to filter out low-speed targets detected by millimeter-wave radar, which prevents it from effectively detecting traffic queues. Therefore, how to leverage video target perception capabilities to compensate for the shortcomings of millimeter-wave radar in static target perception and improve its target perception accuracy has become a problem that needs to be solved. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a radar-visual fusion method based on static target correction. This method overcomes the limitations of millimeter-wave radar in sensing static or low-speed targets. By comparing and analyzing targets detected in video, it identifies potential stationary or quasi-stationary targets that the millimeter-wave radar has not detected. At the same time, it corrects quasi-static targets that the millimeter-wave radar has not identified, thereby improving the sensing accuracy of the millimeter-wave radar.

[0005] The technical solution to achieve the purpose of this invention is: a radar-visual fusion method based on static target correction, the method comprising the following steps:

[0006] Acquire millimeter-wave radar target tracking results and video target detection results;

[0007] The target's motion state is determined based on millimeter-wave radar target tracking results and video target detection results;

[0008] The target's motion state and historical trajectory are used to correct the millimeter-wave radar's target tracking.

[0009] The newly identified targets by the millimeter-wave radar are re-matched, duplicate targets are removed, and new targets are added.

[0010] Update millimeter-wave radar target tracking results.

[0011] Furthermore, the millimeter-wave radar target tracking results include the target's position coordinates, target velocity, target direction of motion, target identifier ID, and whether the current frame successfully identified the target; the video target detection results include the target's position bounding box, target category, and target confidence level.

[0012] Furthermore, the target motion state includes motion, rest, quasi-rest, and disappearance states.

[0013] Furthermore, the step of determining the target motion state based on the millimeter-wave radar target tracking results and video target detection results specifically includes:

[0014] Obtain N millimeter-wave radar target tracking results and M video target detection results. The nth millimeter-wave radar target tracking result is denoted as: (x n y n velo n dirx n diry n id n , detected n The detection result of the m-th video object is denoted as: (bx m by m bw m bh m c m P m ); where x n Let x be the x-coordinate of the nth millimeter-wave radar target, and y be the x-coordinate of the target. n Let y be the y-coordinate of the nth millimeter-wave radar target, velo n For speed, dirx n diry n Let dirx be the direction of target motion. n Let x be the x-component of the motion direction vector of the nth target. n It is the y-component of the motion direction vector of the nth target; id n For the target identifier id, detected n To indicate whether the current frame was successfully identified; bx m by m bw m bh m Let bx be the target location box. m It is the x-coordinate of the center point of the m-th video target bounding box on the video screen, by mm It is the y-coordinate of the center point of the m-th video target bounding box on the video screen, bw m bh is the width of the m-th video target bounding box on the video screen. m c is the height of the m-th video target bounding box in the video frame;m For the target category, P m Target confidence level;

[0015] Iterate through N millimeter-wave radar target tracking results, and perform the following steps for the nth target:

[0016] Calculate the matching degree between each of the M video targets and the nth target, and select the video target bounding box with the highest matching degree, denoted as the kth target, and its matching degree is denoted as match. n，k ;

[0017] Determine the motion state of the nth target based on the following logic:

[0018] If detected n If true, then perform the following judgment:

[0019] If velo n If >motion_threshold, then the target motion state is marked as motion, where motion_threshold is the set motion threshold;

[0020] If velo n If <stationary_threshold, the target's motion state is marked as stationary, where stationary_threshold is the set stationary threshold;

[0021] Otherwise, mark the target's motion state as quasi-stationary;

[0022] If detected n If it is false, then the following judgment is made:

[0023] If match n，k If <match_threshold, the target's motion state is marked as disappeared, where match_threshold is the set disappearance threshold;

[0024] If match n，k If the IoU of the video target bounding box matched at the current time T and the video target bounding box matched at time T-dT is less than IoU_threshold, then the target motion state is marked as stationary; where IoU_threshold is the set IoU threshold and dT is the selected time threshold.

[0025] Otherwise, mark the target's motion state as quasi-static.

[0026] Furthermore, the matching degree between the nth millimeter-wave radar-tracked target and the mth video target is calculated using the following steps:

[0027] Transform the millimeter-wave radar coordinates to the video image coordinate system, denoted as (X... n Y n );

[0028] Calculate the similarity in the x-direction using sim. x The formula is as follows:

[0029]

[0030]

[0031] Calculate the similarity in the y-direction using the sim algorithm. y The formula is as follows:

[0032]

[0033]

[0034] The formula for calculating the total similarity (sim) in the x and y directions is as follows:

[0035]

[0036] in,

[0037] sum = (log x +log y )

[0038]

[0039]

[0040] Add target confidence correction, as shown in the following formula:

[0041] match n，m =p m *sim

[0042] match n，m This represents the matching degree between the nth millimeter-wave radar-tracked target and the mth video target.

[0043] Furthermore, the step of using the target's motion state and historical trajectory to correct the millimeter-wave radar's target tracking specifically includes: correcting targets in stationary and quasi-stationary states, with the following steps:

[0044] If the millimeter-wave radar detects the target in the current frame, the tracking target is updated using the target detection data from that frame;

[0045] If the target is stationary and the millimeter-wave radar does not detect the target in the current frame, the target data from the previous frame is used as the target detection data for the current frame to update the tracking target.

[0046] If the target is in a quasi-static state and the millimeter-wave radar does not detect the target in the current frame, the target detection data for the current frame is estimated based on the target data of the previous frame, and the tracking target is updated using the estimated data.

[0047] Furthermore, the process of estimating the target detection data of the current frame based on the target data of the previous frame is as follows:

[0048] Let the target data of the previous frame be (x) n-1 y n-1 velo n-1 dirx n-1 diry n-1 id n-1 , detected n-1 );

[0049] The target data for the current frame is estimated as follows:

[0050] x n =x n-1 +dirx n-1 *velo n-1 *dt

[0051] y n =y n-1 +diry n-1 *velo n-1 *dt

[0052] velo n =velo n-1

[0053] dirx n =dirx n-1

[0054] diry n =diry n-1

[0055] Where dt is the time interval between two adjacent frames.

[0056] Furthermore, the re-matching of newly identified targets by the millimeter-wave radar, removing duplicate targets and adding newly identified targets, specifically includes:

[0057] The targets tracked by the millimeter-wave radar are divided into P known targets and Q newly identified targets; among them, those that did not exist before but are newly identified in the current frame are marked as newly identified targets, and the rest are marked as known targets.

[0058] The newly identified targets are deduplicated based on their positions and velocities (Q new targets and P known targets).

[0059] The newly identified targets remaining after deduplication are added to the known targets to form the final set of tracked targets in the current frame.

[0060] Furthermore, the deduplication process for the newly identified targets specifically includes:

[0061] Iterate through the Q newly identified targets and the P known targets;

[0062] For the q-th newly identified target and the p-th known target, perform the following steps:

[0063] The distance difference dx and velocity difference dv between the q-th newly identified target and the p-th known target are calculated using the following formulas:

[0064]

[0065] dv=|velo q -velo p |

[0066] Among them, (x q y q ), (x p y p ) are the coordinates of the q-th newly identified target and the p-th known target, respectively, velo q velo p These are the velocities of the q-th newly identified target and the p-th known target, respectively.

[0067] If dx < dist_threshold and dv < velo_threshold, then the newly identified target is determined to be a duplicate target and removed; otherwise, the newly identified target is retained.

[0068] Compared with the prior art, the significant advantages of this invention are:

[0069] 1) Traditional radar-visual fusion schemes generally employ classic Kalman filtering to fuse radar and visual data, but they do not fully consider the characteristics of millimeter-wave radar and video perception, especially the insufficient perception capability of millimeter-wave radar for low-speed targets. This leads to inaccuracies in the fusion results for low-speed targets and an inability to effectively identify target queuing situations. This invention proposes a radar-visual fusion method based on static target correction, which can effectively solve the above problems.

[0070] 2) The present invention proposes a method for identifying target states, which divides the target states into moving, stationary, quasi-stationary and disappearing states. It also proposes to use millimeter-wave radar to track targets and video to detect targets to identify target states, which can effectively identify stationary and quasi-stationary targets that cannot be detected by millimeter-wave radar and correct them.

[0071] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0072] Figure 1 This is a flowchart of a radar-visual fusion method based on static target correction in one embodiment.

[0073] Figure 2 This is a flowchart of target state recognition in one embodiment.

[0074] Figure 3 This is a flowchart of the target correction process in one embodiment.

[0075] Figure 4 This is a flowchart of the new target processing in one embodiment. Detailed Implementation

[0076] 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.

[0077] 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.

[0078] In one embodiment, a radar-visual fusion method based on static target correction is provided, such as... Figure 1 As shown. The specific steps are as follows:

[0079] Step 1: Obtain the millimeter-wave radar target tracking results and video target detection results. The millimeter-wave radar target tracking results include the target's position coordinates, target velocity, target movement direction, target identifier ID, and whether the target was successfully identified in the current frame. The video target detection results include the target's position bounding box, target category, and target confidence level.

[0080] Step 2: Determine the target's motion state based on the millimeter-wave radar target tracking results and video target detection results. The target's motion state includes moving, stationary, quasi-stationary, and vanished states, such as... Figure 2 As shown. The specific steps are as follows:

[0081] Step 2-1: Obtain N millimeter-wave radar target tracking results and M video target detection results. The nth millimeter-wave radar target tracking result is denoted as: (x n y n velo n dirx n diry n id n , detected n The detection result of the m-th video object is denoted as: (bx m by m bw m bh m c m p m ); where x n Let x be the x-coordinate of the nth millimeter-wave radar target, and y be the x-coordinate of the target. n Let y be the y-coordinate of the nth millimeter-wave radar target, velo n For speed, dirx n diry n Let dirx be the direction of target motion. n Let x be the x-component of the motion direction vector of the nth target. n It is the y-component of the motion direction vector of the nth target; id n For the target identifier id, detected n To indicate whether the current frame was successfully identified; bx m by m bw m bh m Let bx be the target location box. m It is the x-coordinate of the center point of the m-th video target bounding box on the video screen, by m It is the y-coordinate of the center point of the m-th video target bounding box on the video screen, bw m bh is the width of the m-th video target bounding box on the video screen. m c is the height of the m-th video target bounding box in the video frame; m For the target category, p m Target confidence level;

[0082] Step 2-2: Traverse the target tracking results of N millimeter-wave radars;

[0083] Steps 2-3: For the nth target, calculate the matching degree between the M video targets and the nth target, and select the video target bounding box with the highest matching degree, denoted as the kth target, and its matching degree is denoted as match. n，k ;

[0084] The steps for calculating the matching degree are as follows:

[0085] Step 2-3-1: Transform the millimeter-wave radar coordinates to the video image coordinate system, denoted as X. n Y n ;

[0086] Step 2-3-2: Calculate the similarity sim in the x-direction. x The formula is as follows:

[0087]

[0088]

[0089] Step 2-3-3: Calculate the similarity sim in the y-direction. y The formula is as follows:

[0090]

[0091]

[0092] Steps 2-3-4: Calculate the total similarity in the x and y directions, using the following formula:

[0093]

[0094] in,

[0095] sum = (log x +log y )

[0096]

[0097]

[0098] Steps 2-3-5: Add target confidence correction, the formula is as follows:

[0099] match n，m =p m *sim

[0100] Steps 2-4: Determine the motion state of the nth target based on the following logic:

[0101] If detected n If true, then perform the following judgment:

[0102] If velo n If >motion_threshold, then the target motion state is marked as motion, where motion_threshold is the set motion threshold;

[0103] If velo n If <stationary_threshold, the target's motion state is marked as stationary, where stationary_threshold is the set stationary threshold;

[0104] Otherwise, mark the target's motion state as quasi-stationary.

[0105] If detected n If it is false, then the following judgment is made:

[0106] If match n，k If <match_threshold, the target's motion state is marked as disappeared, where match_threshold is the set disappearance threshold;

[0107] If match n，k If the IoU of the video target bounding box matched at the current time T and the video target bounding box matched at time T-dT is less than IoU_threshold, then the target motion state is marked as stationary; where IoU_threshold is the set IoU threshold and dT is the selected time threshold.

[0108] Otherwise, mark the target's motion state as quasi-static.

[0109] Step 3: Correct the millimeter-wave radar target tracking by utilizing the target's motion state and historical trajectory, such as... Figure 3 As shown. The steps are as follows:

[0110] Step 3-1: If the millimeter-wave radar detects the target in the current frame, update the tracking target with the target detection data of that frame;

[0111] Step 3-2: If the target is stationary and the millimeter-wave radar does not detect the target in the current frame, then update the tracking target using the target data from the previous frame as the target detection data for the current frame;

[0112] Step 3-3: If the target is in a quasi-static state and the millimeter-wave radar does not detect the target in the current frame, then estimate the target detection data for the current frame based on the target data of the previous frame, and update the tracking target with the estimated data.

[0113] The target detection data for the current frame is estimated based on the target data of the previous frame, as follows:

[0114] Let the target data of the previous frame be (x) n-1 y n-1 velo n-1 dirx n-1 diry n-1 id n-1 , detected n-1 );

[0115] The target data for the current frame is estimated as follows:

[0116] x n =x n-1 +dirx n-1 *velo n-1 *dt

[0117] y n =y n-1 +diry n-1 *velo n-1 *dt

[0118] velo n =velo n-1

[0119] dirx n =dirx n-1

[0120] diry n =diry n-1

[0121] Where dt is the time interval between two adjacent frames.

[0122] Step 4: Re-match newly identified targets by the millimeter-wave radar, remove duplicate targets, and add newly identified targets, such as... Figure 4 As shown, the specific steps are as follows:

[0123] Step 4-1: Divide the targets tracked by the millimeter-wave radar into P known targets and Q newly identified targets. Targets that did not exist before and are newly identified in the current frame are recorded as newly identified targets, and the rest are recorded as known targets.

[0124] Step 4-2: Add the newly identified targets remaining after deduplication to the known targets, thus forming the final set of tracked targets for the current frame.

[0125] The specific steps for deduplication are as follows:

[0126] Step 4-2-1: Iterate through the Q newly identified targets and the P known targets;

[0127] Step 4-2-2: For the q-th newly identified target and the p-th known target, perform the following steps:

[0128] Step 4-2-3: Calculate the distance and velocity difference between the q-th newly identified target and the p-th known target, using the following formula:

[0129]

[0130] dv=|velo q -velo p |

[0131] Step 4-2-4: If dx < dist_threshold and dv < velo_threshold, then the new target is determined to be a duplicate target and the new target is removed; otherwise, the new target is retained.

[0132] Step 5: Update the millimeter-wave radar tracking results.

[0133] In one embodiment, a radar-visual fusion system based on static target correction is provided, the system comprising:

[0134] The first module is used to acquire millimeter-wave radar target tracking results and video target detection results;

[0135] The second module is used to determine the target's motion state based on the millimeter-wave radar target tracking results and video target detection results;

[0136] The third module is used to correct the millimeter-wave radar's tracking of the target by utilizing the target's motion state and historical trajectory.

[0137] The fourth module is used to re-match newly identified targets by the millimeter-wave radar, remove duplicate targets, and add newly identified targets.

[0138] The fifth module is used to update the target tracking results of the millimeter-wave radar.

[0139] Specific limitations regarding the radar-visual fusion system based on static target correction can be found in the limitations of the radar-visual fusion method based on static target correction mentioned above, and will not be repeated here. Each module in the aforementioned radar-visual fusion system based on static target correction 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.

[0140] 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, when executing the computer program, implements:

[0141] Acquire millimeter-wave radar target tracking results and video target detection results;

[0142] The target's motion state is determined based on millimeter-wave radar target tracking results and video target detection results;

[0143] The target's motion state and historical trajectory are used to correct the millimeter-wave radar's target tracking.

[0144] The newly identified targets by the millimeter-wave radar are re-matched, duplicate targets are removed, and new targets are added.

[0145] Update millimeter-wave radar target tracking results.

[0146] For specific limitations on each step, please refer to the limitations of the radar-visual fusion method based on static target correction mentioned above, which will not be repeated here.

[0147] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being implemented when executed by a processor:

[0148] Acquire millimeter-wave radar target tracking results and video target detection results;

[0149] The target's motion state is determined based on millimeter-wave radar target tracking results and video target detection results;

[0150] The target's motion state and historical trajectory are used to correct the millimeter-wave radar's target tracking.

[0151] The newly identified targets by the millimeter-wave radar are re-matched, duplicate targets are removed, and new targets are added.

[0152] Update millimeter-wave radar target tracking results.

[0153] For specific limitations on each step, please refer to the limitations of the radar-visual fusion method based on static target correction mentioned above, which will not be repeated here.

[0154] 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 radar-visual fusion method based on static target correction, characterized in that, The method includes the following steps: Acquire millimeter-wave radar target tracking results and video target detection results; The target's motion state is determined based on millimeter-wave radar target tracking results and video target detection results; The target's motion state and historical trajectory are used to correct the millimeter-wave radar's target tracking. The newly identified targets by the millimeter-wave radar are re-matched, duplicate targets are removed, and new targets are added. Update millimeter-wave radar target tracking results; The determination of target motion state based on millimeter-wave radar target tracking results and video target detection results specifically includes: Obtain N millimeter-wave radar target tracking results and M video target detection results. The nth millimeter-wave radar target tracking result is denoted as: The detection result of the m-th video object is denoted as: ;in, for , For speed magnitude, Let be the direction of the target motion, where The direction vector of the nth target's motion Quantity, It is the first The target motion direction vector Quantity; For the target identifier id, To determine whether the current frame was successfully identified; The target location is defined by the bounding box, where It is the center point of the m-th video target bounding box on the video screen. coordinate, It is the y-coordinate of the center point of the m-th video target bounding box on the video screen. It is the width of the bounding box of the m-th video target on the video screen. It is the height of the m-th video target's bounding box in the video frame; For the target category, Target confidence level; Iterate through N millimeter-wave radar target tracking results, and perform the following steps for the nth target: Calculate the matching degree between each of the M video targets and the nth target, and select the video target bounding box with the highest matching degree, denoted as the kth target, and its matching degree is denoted as . ; Determine the motion state of the nth target based on the following logic: like If true, then perform the following judgment: if Then the target's motion state is marked as motion, where The set motion threshold; if If the target's motion state is marked as stationary, then... The set static threshold; Otherwise, mark the target's motion state as quasi-stationary; like If it is false, then the following judgment is made: if If the target's motion state is marked as vanished, then... The set disappearance threshold; if Furthermore, the inou (iou) between the video target bounding box matched at the current time T and the video target bounding box matched at time T-dT is less than 1. If the target's motion state is marked as stationary, then the target's motion state is marked as stationary; where dT is the set IOU threshold, and dT is the selected time threshold; Otherwise, mark the target's motion state as quasi-static; The steps to calculate the matching degree between the nth millimeter-wave radar-tracked target and the mth video target are as follows: Transform the millimeter-wave radar coordinates to the video image coordinate system, denoted as... ; Calculate the similarity in the x-direction. The formula is as follows: Calculate the similarity in the y-direction. The formula is as follows: Calculate the total similarity in the x and y directions. The formula is as follows: in, Add target confidence correction, as shown in the following formula: This represents the matching degree between the nth millimeter-wave radar-tracked target and the mth video target.

2. The radar-visual fusion method based on static target correction according to claim 1, characterized in that, The millimeter-wave radar target tracking results include the target's position coordinates, target velocity, target direction of motion, target identifier ID, and whether the current frame successfully identified the target; the video target detection results include the target's location bounding box, target category, and target confidence level.

3. The radar-visual fusion method based on static target correction according to claim 1, characterized in that, The target motion state includes motion, stillness, quasi-stillness, and disappearance.

4. The radar-visual fusion method based on static target correction according to claim 3, characterized in that, The step of correcting the millimeter-wave radar tracking of targets using the target's motion state and historical trajectory specifically includes: correcting targets in stationary and quasi-stationary states, with the following steps: If the millimeter-wave radar detects the target in the current frame, the tracking target is updated using the target detection data from that frame; If the target is stationary and the millimeter-wave radar does not detect the target in the current frame, the target data from the previous frame is used as the target detection data for the current frame to update the tracking target. If the target is in a quasi-static state and the millimeter-wave radar does not detect the target in the current frame, the target detection data for the current frame is estimated based on the target data of the previous frame, and the tracking target is updated using the estimated data.

5. The radar-visual fusion method based on static target correction according to claim 4, characterized in that, The process of estimating the target detection data of the current frame based on the target data of the previous frame is as follows: Let the target data of the previous frame be... ;in, for , for Speed ​​magnitude, for The direction of target movement, where, For the first -1 target motion direction vector Quantity, It is the first The target motion direction vector Quantity; To determine whether the previous frame was successfully identified; The target data for the current frame is estimated as follows: Where dt is the time interval between two adjacent frames.

6. The radar-visual fusion method based on static target correction according to claim 1, characterized in that, The process of re-matching newly identified targets by millimeter-wave radar, removing duplicate targets, and adding newly identified targets specifically includes: The targets tracked by the millimeter-wave radar are divided into P known targets and Q newly identified targets; among them, targets that did not exist before and are newly identified in the current frame are recorded as newly identified targets, and the rest are recorded as known targets; The newly identified targets are deduplicated based on their positions and velocities (Q new targets and P known targets). The newly identified targets remaining after deduplication are added to the known targets to form the final set of tracked targets in the current frame.

7. The radar-visual fusion method based on static target correction according to claim 6, characterized in that, The deduplication process for newly identified targets specifically includes: Iterate through the Q newly identified targets and the P known targets; For the q-th newly identified target and the p-th known target, perform the following steps: Calculate the distance difference between the q-th newly identified target and the p-th known target. and speed difference The formula is as follows: in, ), ) are the coordinates of the q-th newly identified target and the p-th known target, respectively. These are the velocities of the q-th newly identified target and the p-th known target, respectively. if and If the newly identified target is found to be a duplicate target, it is removed; otherwise, the newly identified target is retained.

8. A radar-visual fusion system based on static target correction according to the method of any one of claims 1 to 7, the system comprising: The first module is used to acquire millimeter-wave radar target tracking results and video target detection results; The second module is used to determine the target's motion state based on the millimeter-wave radar target tracking results and video target detection results; The third module is used to correct the millimeter-wave radar's tracking of the target by utilizing the target's motion state and historical trajectory. The fourth module is used to re-match newly identified targets by the millimeter-wave radar, remove duplicate targets, and add newly identified targets. The fifth module is used to update the target tracking results of the millimeter-wave radar.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images