Detection Method, Device, Equipment, Storage Medium and Vehicle for Static Targets

Abstract

The present invention relates to the technical field of target detection, and discloses a method, device, equipment, storage medium and vehicle for detecting stationary targets. The method includes: acquiring a current frame image collected by a camera, and identifying a camera observation target in the current frame image; associating the camera observation target with a current target detection frame, and performing state prediction update on the associated target detection frame according to the relative motion parameters of the camera observation target; the target detection frame includes a detection frame created based on historical targets observed; determining whether the state of the updated target detection frame converges; and determining whether the updated target detection frame is stationary in the case where the state of the updated target detection frame converges. The present invention associates an observation target with a target detection frame and performs state update, and by using the target detection frame with converged state, it can more accurately determine whether the target detection frame is stationary.

Description

Technical Field

[0001] The present invention relates to the technical field of target detection, and specifically to a method, device, equipment, storage medium and vehicle for detecting stationary targets. Background Art

[0002] Intelligent driving vehicles are equipped with various sensors such as cameras and radars. Based on the data collected by the sensors, objects around the vehicle can be detected to enable avoidance. Among them, stationary objects have a great impact on vehicle driving, and it is necessary to accurately detect stationary objects around the vehicle.

[0003] When detecting stationary targets, especially for stationary targets at a long distance (for example, the distance exceeds 100m), the speed measurement and ranging accuracy of cameras is not high, and radars are prone to false detection of stationary targets. Some solutions fuse the image features and radar features of cameras, but data fusion requires a large amount of calculation and the processing efficiency is low. Summary of the Invention

[0004] In view of this, the present invention provides a method, device, equipment, storage medium and vehicle for detecting stationary targets to solve the problem of inaccurate detection of stationary targets.

[0005] In a first aspect, the present invention provides a method for detecting a stationary target, including:

[0006] Obtaining a current frame image collected by a camera and identifying a camera observation target in the current frame image;

[0007] Associating the camera observation target with a current target detection frame, and updating the state prediction of the associated target detection frame according to the relative motion parameters of the camera observation target; the target detection frame includes a detection frame created based on observed historical targets;

[0008] Determining whether the updated target detection frame converges in state;

[0009] When the state of the updated target detection frame converges, determining whether the updated target detection frame is stationary.

[0010] In some optional embodiments, the associating the camera observation target with the current target detection frame includes:

[0011] Associating a first camera observation target with the current target detection frame; the first camera observation target is the camera observation target corresponding to the most critical driving object;

[0012] After that, associate the second camera observation target with the target detection box that is not currently associated with the first camera observation target; the second camera observation target is other camera observation targets except the first camera observation target;

[0013] Create a new target detection box for the second camera observation target that fails to be associated with the target detection box.

[0014] In some optional embodiments, the associating the camera observation target with the current target detection box further includes:

[0015] In the case where the first camera observation target fails to be associated with the target detection box, after creating a new target detection box for the second camera observation target that fails to be associated with the target detection box, associate the first camera observation target with the target detection box that is not currently associated with the camera observation target again;

[0016] In the case where the first camera observation target fails to be associated with the target detection box again, create a new target detection box for the first camera observation target.

[0017] In some optional embodiments, the determining whether the state of the updated target detection box converges includes:

[0018] Determine the first association times of the updated target detection box; the first association times is the number of times the updated target detection box associates with the camera observation target corresponding to the most critical moving object;

[0019] In the case where the first association times is greater than the first threshold, determine that the state of the updated target detection box converges;

[0020] And / or,

[0021] Determine the second association times of the updated target detection box; the second association times is the number of times the updated target detection box associates with the camera observation target;

[0022] In the case where the second association times is greater than the second threshold, determine the covariance matrix between the relative speed and the relative depth corresponding to the updated target detection box;

[0023] In the case where the covariance matrix satisfies the convergence condition, determine that the state of the updated target detection box converges.

[0024] In some optional embodiments, the covariance matrix is:

[0025]

[0026] where depth represents the relative depth measurement, speed represents the relative speed measurement, represents the covariance matrix between the relative speed and the relative depth of P(depth, speed), and P dd represents the variance of the relative depth measurement, and P ss represents the variance of the relative speed measurement, and P ds represents the covariance between the relative depth measurement and the relative speed measurement, and P sd represents the covariance between the relative speed measurement and the relative depth measurement;

[0027] The convergence condition includes:

[0028]

[0029] where depth′ is the relative depth of the target detection box, and both k1 and k2 are preset threshold parameters.

[0030] In some alternative embodiments, when determining whether the updated target detection box is stationary under the condition that the state of the updated target detection box converges, it includes:

[0031] When the state of the updated target detection box converges, if the relative speed of the updated target detection box is less than the current speed threshold, it is determined that the updated target detection box is stationary;

[0032] Or,

[0033] When the state of the updated target detection box converges, taking the negative value of its own speed as a variable and the relative speed of the updated target detection box as the expectation, determine the cumulative probability of the normal distribution corresponding to the own speed;

[0034] When the cumulative probability of the normal distribution is greater than the preset probability value, it is determined that the updated target detection box is stationary.

[0035] In some alternative embodiments, the cumulative probability of the normal distribution is:

[0036]

[0037] where x represents the negative value of the own speed, Δv represents the preset speed difference, μ represents the relative speed expectation, and σ represents the relative speed standard deviation.

[0038] In some alternative embodiments, the method further includes:

[0039] Determine the radar observation targets currently identified by the radar and filter out the stationary radar observation targets;

[0040] Associate the stationary radar observation target with the current target detection box, and update the state prediction of the associated target detection box according to the relative motion parameters of the stationary radar observation target.

[0041] In some alternative embodiments, determining whether the state of the updated target detection box converges includes:

[0042] Determine the third association count of the updated target detection box; the third association count is the number of times the updated target detection box is associated with the stationary radar observation target; in the case where the third association count is greater than a third threshold, determine that the state of the updated target detection box converges;

[0043] In the case where the state of the updated target detection box converges, determining whether the updated target detection box is stationary includes:

[0044] In the case where the third association count of the updated target detection box associated with the stationary radar observation target is greater than the third threshold, determine that the updated target detection box is stationary.

[0045] In some alternative embodiments, the method further includes:

[0046] Delete the target detection boxes that meet the failure conditions;

[0047] Wherein, the failure conditions include: the duration that the target detection box is not associated exceeds a first preset duration; or, the target type of the target detection box is inconsistent with the target type of the associated observation target, and the inconsistent duration exceeds a second preset duration.

[0048] In a second aspect, the present invention provides a detection device for stationary targets, including:

[0049] A target recognition module, configured to obtain the current frame image collected by the camera and recognize the camera observation target in the current frame image;

[0050] An association update module, configured to associate the camera observation target with the current target detection box, and update the state prediction of the associated target detection box according to the relative motion parameters of the camera observation target; the target detection box includes a detection box created based on the observed historical targets;

[0051] A convergence judgment module, configured to judge whether the state of the updated target detection box converges;

[0052] A detection module, configured to determine whether the updated target detection box is stationary in the case where the state of the updated target detection box converges.

[0053] In a third aspect, the present invention provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the method for detecting a stationary target according to the first aspect or any corresponding embodiment thereof as described above.

[0054] In a fourth aspect, the present invention provides a computer-readable storage medium, on which computer instructions are stored. The computer instructions are used to cause a computer to execute the method for detecting a stationary target according to the first aspect or any corresponding embodiment thereof as described above.

[0055] In a fifth aspect, the present invention provides a vehicle, including a vehicle controller, and the vehicle controller is used to execute the method for detecting a stationary target according to the first aspect or any corresponding embodiment thereof as described above.

[0056] The present invention associates the observed target observed by the camera with the target detection frame and updates the state. After multiple rounds of processing, the state of the target detection frame can converge. Even if the observed value of the observed target has a large observation error, the target detection frame with converged state can be used to more accurately represent the true state of the corresponding observed target, reduce the influence caused by the observation error, and thus can more accurately determine whether the target detection frame is stationary, and has a good detection effect on stationary targets at a long distance, with high accuracy and recall rate; when applied to an autonomous vehicle, it can accurately identify stationary objects such as obstacles in the lane, and improve the safety of the autonomous vehicle. Detecting stationary targets based on camera vision has low cost, does not require data fusion, has a small amount of calculation, and high recognition efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the related art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0058] Figure 1 is a flowchart of the method for detecting a stationary target according to an embodiment of the present invention;

[0059] Figure 2 is a flowchart of another method for detecting a stationary target according to an embodiment of the present invention;

[0060] Figure 3 is a schematic diagram of a process for associating a camera observed target according to an embodiment of the present invention;

[0061] Figure 4 It is a schematic flowchart of another method for detecting stationary objects according to an embodiment of the present invention;

[0062] Figure 5 It is a structural block diagram of a device for detecting stationary objects according to an embodiment of the present invention;

[0063] Figure 6 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Specific embodiments

[0064] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0065] During the process of autonomous driving or assisted driving of a vehicle, it is necessary to use sensors such as cameras and radars installed on the vehicle to detect objects around the vehicle to remind the driver to avoid or automatically avoid them. Since stationary objects have a great impact on the safe driving of the vehicle, it is particularly important to be able to accurately detect stationary objects.

[0066] For example, when vehicle A is in the process of autonomous driving and there is another vehicle B in front of it. If vehicle B is temporarily parked in the middle of the lane due to reasons such as a breakdown or a traffic accident, at this time, the camera, radar, etc. in vehicle A need to be able to detect vehicle B in front and determine that it is currently in a stationary state, that is, vehicle B is a stationary target. Thus, vehicle A can automatically avoid vehicle B and prevent a collision. However, if vehicle A misidentifies vehicle B as non-stationary, for example, misidentifies that vehicle B is also moving forward, then vehicle A may not be able to avoid vehicle B in time due to misjudgment, thus there is a risk of collision.

[0067] In addition, when sensors such as cameras and radars detect the state of distant objects, there are generally large errors, and it is difficult to accurately determine information such as the relative speed and relative position of distant objects. If stationary objects in the distance can be accurately identified, it can enable vehicle A to avoid stationary objects in advance when driving at high speed on road conditions such as highways, ensuring the safety of vehicle driving.

[0068] According to an embodiment of the present invention, an embodiment of a method for detecting stationary objects is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0069] In this embodiment, a method for detecting stationary targets is provided, which can be applied to local terminals such as vehicle controllers, or to servers located on the network side, or to distributed systems located on the network side, specifically depending on the actual situation. Figure 1 It is a flowchart of the method for detecting stationary targets according to an embodiment of the present invention, as Figure 1 shown, and the process includes the following steps.

[0070] Step S101, obtain the current frame image collected by the camera, and identify the camera observation targets in the current frame image.

[0071] In this embodiment, when detecting stationary targets, the camera is used to collect images in real time. A frame of image collected by the camera at the current moment is called the current frame image; among them, the current frame image includes the targets around the camera, and these targets are observed by the camera and are hereinafter referred to as camera observation targets.

[0072] For example, the camera is a vehicle-mounted camera, which can collect the surrounding environment images of the vehicle in real time, generate the corresponding current frame image, and by performing target detection on the current frame image, targets such as vehicles and pedestrians in the current frame image can be identified, and these targets are all camera observation targets. Among them, some targets may be stationary targets. For example, the stationary targets can be vehicles and pedestrians stopped in the lane, or roadblocks, falling stones in the lane, etc.

[0073] Step S102, associate the camera observation targets with the current target detection boxes, and update the state prediction of the associated target detection boxes according to the relative motion parameters of the camera observation targets; the target detection boxes include the detection boxes created based on the observed historical targets.

[0074] In this embodiment, when performing target detection, corresponding detection boxes will be created for the observed targets, and these detection boxes can be used as target detection boxes; among them, the target detection boxes are the detection boxes corresponding to the previously observed historical targets. Corresponding target detection boxes can be created for the initially observed targets; for example, based on the first frame image collected by the camera, the observed targets therein can be identified, and initial target detection boxes can be created for each observed target, and subsequently, based on these initial target detection boxes, they are associated with the camera observation targets.

[0075] Specifically, after determining the current camera observation targets, these camera observation targets can be associated with the current target detection boxes to determine which target detection boxes the camera observation targets are associated with. For example, by calculating the IOU (Intersection over Union) or size ratio between the camera observation targets and the target detection boxes, it can be determined which target detection boxes the camera observation targets can be associated with.

[0076] Moreover, each camera-observed target has certain relative motion parameters. If a camera-observed target 1 is associated with a target detection box A, the state of the target detection box A can be predicted based on the relative motion parameters of the camera-observed target 1, so as to update the state of the target detection box A. For example, the relative motion parameters may specifically include a relative velocity (i.e., the relative velocity between itself and the camera-observed target) and a relative depth (i.e., the relative depth between itself and the camera-observed target, which can also be referred to as a relative distance). The state of the target detection box A can be predicted and updated based on methods such as Kalman filtering and particle filtering to update the state information such as the position, velocity, and acceleration of the target detection box A, which will not be elaborated here.

[0077] For example, after creating an initial target detection box based on the first frame image, since the initial target detection box corresponds one-to-one with the observed targets, the state of the initial target detection box can be predicted and updated based on the relative motion parameters of these observed targets. These updated target detection boxes can then be associated with the camera-observed targets in the second frame image, that is, the second frame image is the current frame image, and the above step S102 is executed. Repeating this process can achieve real-time association and update of the target detection box.

[0078] Step S103, determine whether the state of the updated target detection box converges.

[0079] In this embodiment, whenever a current frame image is acquired, the target detection box can be associated and updated based on the currently identified camera-observed targets. After multiple rounds of updates, the state of the target detection box can tend to be stable, that is, the state converges.

[0080] Taking the Kalman filter state prediction of the target detection box as an example, the more times the target detection box is updated, the more stable its state becomes. Therefore, it can be determined whether the state converges based on the number of updates of the target detection box.

[0081] Step S104, when the state of the updated target detection box converges, determine whether the updated target detection box is stationary.

[0082] If the state of the target detection box converges, it means that the state (position, velocity, acceleration, etc.) of the target detection box has tended to be stable. At this time, even if there is a large error in the camera-observed target, for example, there is a large error in the velocity observation value of the camera-observed target, and the target detection box is updated with this velocity observation value with a large error, it can still ensure that the velocity of the target detection box is relatively close to the true velocity of the camera-observed target. In other words, for a target detection box with a converged state, its state is stable and is less affected by the error of the observation value.

[0083] Therefore, the object detection frame based on state convergence can relatively accurately represent the state of the corresponding observed object, reduce the impact caused by observation errors. Especially for long-distance object detection, it can reduce the problem of inaccurate measurement caused by long distance and large errors, and thus can accurately determine whether the object detection frame is stationary. If the object detection frame is stationary, the observed object associated with the object detection frame is also stationary. At this time, the stationary object detection frame can be sent downstream for planning and control.

[0084] The method for detecting stationary objects provided by the embodiments of the present invention associates the observed objects observed by the camera with the object detection frame and updates the state. After multiple rounds of processing, the state of the object detection frame can converge. Even if there are large observation errors in the observed values of the observed objects, the object detection frame with converged state can be used to accurately represent the true state of the corresponding observed object, reduce the impact caused by observation errors, and thus can more accurately determine whether the object detection frame is stationary, and has a good detection effect on stationary objects at a long distance, with high accuracy and recall rate; when applied to an autonomous vehicle, it can accurately identify stationary objects such as obstacles in the lane, and improve the safety of the autonomous vehicle. Detecting stationary objects based on camera vision has low cost, does not require data fusion, has a small amount of calculation, and high recognition efficiency.

[0085] In this embodiment, a method for detecting stationary objects is provided, which can be applied to local terminals such as vehicle controllers, or to servers located on the network side, or to distributed systems located on the network side. Figure 2 It is a flowchart of the method for detecting stationary objects according to the embodiments of the present invention, as Figure 2 shown, and this process includes the following steps.

[0086] Step S201, obtain the current frame image collected by the camera, and identify the camera observed object in the current frame image.

[0087] For details, please refer to Figure 1 step S101 of the embodiment shown, which will not be elaborated here.

[0088] Step S202, associate the camera observed object with the current object detection frame, and perform state prediction update on the associated object detection frame according to the relative motion parameters of the camera observed object; the object detection frame includes the detection frame created based on the observed historical objects.

[0089] For details, please refer to Figure 1 step S102 of the embodiment shown, which will not be elaborated here.

[0090] In some alternative embodiments, the above step S202, "associating the camera observation target with the current target detection box", includes the following steps A1 to A3.

[0091] Step A1, associating the first camera observation target with the current target detection box; the first camera observation target is the camera observation target corresponding to the most critical driving object.

[0092] Step A2, then, associating the second camera observation target with the target detection box that has not been associated with the first camera observation target; the second camera observation target is the other camera observation targets except the first camera observation target.

[0093] Step A3, creating a new target detection box for the second camera observation target that fails to be associated with the target detection box.

[0094] In the autonomous driving scenario, there is the most critical driving object (Most critical dynamic object, MCDO) that affects vehicle driving. Since the first object in the self-lane has the highest degree of danger, generally the first object in the self-lane is used as the MCDO. After identifying multiple camera observation targets in the current frame image, it is possible to determine which camera observation target is the MCDO observation target. For the sake of description, the camera observation target corresponding to the MCDO is called the first camera observation target, and the other camera observation targets except the first camera observation target are called the second camera observation targets, that is, the second camera observation targets are non-MCDO observation targets.

[0095] When performing detection box association, first associate the first camera observation target related to the MCDO, which can increase the probability of the first camera observation target being associated and avoid the risk of association failure caused by conflicts with other camera observation targets.

[0096] After associating the first camera observation target, the remaining second camera observation targets can be associated. Specifically, associate the second camera observation target with the target detection box that has not been associated with the first camera observation target.

[0097] Among them, if the first camera observation target is successfully associated with a certain target detection box, the second camera observation target cannot be associated with the successfully associated target detection box anymore. If the first camera observation target fails to be associated, the second camera observation target can try to be associated with any target detection box.

[0098] Moreover, if there are second camera observation targets that fail to be associated with the target detection box, new target detection boxes are created for these second camera observation targets; if all the second camera observation targets are successfully associated, no additional processing is required for the second camera observation targets.

[0099] In this embodiment, by first associating the camera observation targets corresponding to MCDO, the camera observation targets corresponding to MCDO can be preferentially associated and updated, which can improve the recall rate of MCDO targets, thereby enhancing the safety of autonomous vehicles. Creating new target detection frames for the second camera observation targets that fail to associate with the target detection frames can improve the detection rate of stationary targets; moreover, creating new target detection frames based on camera observation targets does not require the use of radar, which can avoid the problem of a large number of erroneously created target detection frames caused by radar ghosts.

[0100] Optionally, the above step S202, "associating the camera observation targets with the current target detection frames", may further include the following steps A4 to A5.

[0101] Step A4, in the case where the first camera observation target fails to associate with the target detection frame, after creating a new target detection frame for the second camera observation target that fails to associate with the target detection frame, re-associate the first camera observation target with the target detection frames that are not currently associated with the camera observation targets.

[0102] Step A5, in the case where the first camera observation target fails to associate with the target detection frame again, create a new target detection frame for the first camera observation target.

[0103] In the embodiment of the present invention, if the first camera observation target fails to associate with the target detection frame, then similar to the above step A3, a new target detection frame can also be created for the first camera observation target. Among them, to improve the detection recall rate of MCDO, the first camera observation target is re-associated.

[0104] Specifically, if in the above step A1, the first camera observation target fails to associate with the target detection frame, then after the above step A3, "creating a new target detection frame for the second camera observation target that fails to associate with the target detection frame", the first camera observation target is re-associated again, that is, the first camera observation target is re-associated with the target detection frames that are not currently associated with the camera observation targets. If the first camera observation target is successfully associated, only the status prediction of the associated target detection frame needs to be updated; if the first camera observation target fails to be re-associated again, that is, it fails to associate with the target detection frame again, then a new target detection frame is also created for the first camera observation target.

[0105] Optionally, the method further includes step A6.

[0106] Step A6, deleting the target detection frames that meet the failure conditions.

[0107] Among them, the failure conditions include: the duration for which the target detection box is not associated exceeds a first preset duration; or, the target type of the target detection box is inconsistent with the target type of the associated observed target, and the duration of the inconsistency exceeds a second preset duration.

[0108] In an embodiment of the present invention, for a certain target detection box, if the duration for which it is not associated exceeds a first preset duration (such as 300 ms, 400 ms, etc.), it indicates that the target detection box has failed to be associated for a long time, and the target detection box can be deleted.

[0109] Alternatively, if a certain target detection box has established an association relationship with an observed target, but if the target type of the target detection box is inconsistent with the target type of the associated observed target for a long time, that is, the duration of the inconsistency exceeds a second preset duration (such as 200 ms, 300 ms, etc.), it can also be considered that the association between the two is invalid, and the target detection box can also be deleted. For example, if the target type of a certain target detection box is "vehicle", but the target type of the associated observed target is "pedestrian", if the target types of the two are inconsistent for a long time, the target detection box is deleted.

[0110] Figure 3 Fig. shows a schematic process diagram of associating camera observed targets. As Figure 3 shown, if there are currently seven target detection boxes, namely target detection boxes A - G; and, six camera observed targets are recognized in the current frame image, namely camera observed targets 1 - 6.

[0111] First, MCDO association is performed. As Figure 3 shown, if camera observed target 1 corresponds to MCDO, then the camera observed target 1 is associated with the target detection box; if the IOU between camera observed target 1 and target detection box A is greater than the threshold, then the two can be associated, and subsequently, the state prediction of target detection box A is updated based on the relative motion parameters of the camera observed target 1.

[0112] After that, non - MCDO association is performed. As Figure 3 shown, camera observed targets 2 - 6 are all second camera observed targets, and these camera observed targets can be associated with the remaining target detection boxes. As Figure 3 shown, camera observed targets 2, 3, 4, 5 are respectively associated with target detection boxes B, C, D, E, while camera observed target 6 fails to be associated.

[0113] After that, a new target detection box can be created for the camera observed target that fails to be associated. As Figure 3 shown, based on parameters such as the position, speed, and acceleration of camera observed target 6, a new target detection box H can be created.

[0114] Moreover, if a target detection box meets the failure condition, the failed target detection box is deleted. For example Figure 3 As shown, if the target detection box G has not been associated for a long time, the target detection box G can be deleted. When the next frame of image is collected, the camera observation targets extracted therefrom are used to be associated with the target detection boxes A, B, C, D, E, F, and H.

[0115] It should be noted that in this embodiment, the target detection box can also be associated based on the radar detection result. If a target detection box is not associated with the camera observation target but is successfully associated with the radar detection result, then the target detection box is an associated target detection box; only when the target detection box has not been associated with the observation targets of the camera and the radar for a long time, will the target detection box be deleted.

[0116] In this embodiment, the second-round association update of the first camera observation target corresponding to MCDO can improve the detection recall rate of MCDO and effectively avoid the missed detection of MCDO observation targets; timely deleting the failed target detection boxes can improve the accuracy of target detection.

[0117] Step S203: Determine the radar observation targets currently recognized by the radar and filter out the stationary radar observation targets.

[0118] Since camera detection is greatly affected by light, problems such as missed detection may occur. Based on this, in this embodiment, radar is also combined for target detection.

[0119] Specifically, based on the radar data collected by the radar, the targets observed by the radar can be recognized, that is, the radar observation targets. The radar observation targets can specifically include radar point cloud targets. Moreover, by tracking the radar point cloud targets, corresponding radar track targets can be generated, and the radar track targets can also be used as a kind of radar observation targets.

[0120] For each radar observation target, the probability that it belongs to an obstacle can be determined. For the radar observation targets with a probability exceeding the threshold, it can be initially considered that they are stationary, and these radar observation targets will be called stationary radar observation targets subsequently.

[0121] Among them, the radar can specifically be a millimeter-wave radar, and there is no need to use an expensive lidar, which reduces the hardware cost of the autonomous vehicle. Moreover, camera detection and radar detection can run in parallel without affecting each other, and there is no need for data fusion, with a small amount of calculation and high processing efficiency.

[0122] Step S204: Associate the stationary radar observation targets with the current target detection boxes and update the state prediction of the associated target detection boxes according to the relative motion parameters of the stationary radar observation targets.

[0123] In this embodiment, similar to the identification of the camera observation target, after determining the stationary radar observation target, these stationary radar observation targets are associated with the current target detection frame, and then the state prediction of the associated target detection frame can be updated. Details are not described here.

[0124] Among them, since the radar observation target generally does not distinguish the target type, there are useless targets among the selected stationary radar observation targets, such as fences and trees by the roadside; while the target detection frame is created based on the target observed by the camera, and the vision-based target detection frame has nothing to do with useless targets such as fences and trees. By associating the stationary radar observation target with the current target detection frame, the useless targets identified by the radar can also be excluded.

[0125] It can be understood that the camera target detection and the radar target detection are independent. Therefore, for a certain target detection frame, if it is associated with a certain camera detection target, the state prediction can be updated once based on the relative motion parameters of the camera detection target; if the target detection frame is also associated with a certain stationary radar observation target, the state prediction can be updated again based on the relative motion parameters of the stationary radar detection target.

[0126] Step S205, determine whether the state of the updated target detection frame converges.

[0127] For details, please refer to Figure 1 Step S103 of the illustrated embodiment, which will not be elaborated here.

[0128] In some optional embodiments, for relatively important MCDO observation targets, that is, the first camera observation targets, the convergence of the target detection frame associated with the MCDO observation target can be judged separately. Specifically, step S205 "determine whether the state of the updated target detection frame converges" can specifically include steps B1 to B2.

[0129] Step B1, determine the first association times of the updated target detection frame; the first association times are the times when the updated target detection frame is associated with the camera observation target corresponding to the most critical moving object.

[0130] Step B2, when the first association times are greater than the first threshold, determine that the state of the updated target detection frame converges.

[0131] In this embodiment, as described above, the camera observation targets include the first camera observation target corresponding to the most critical moving object and the second camera observation target other than the first camera observation target. For any target detection box, if it is associated with the first camera observation target, the number of times the target detection box is associated with the camera observation target corresponding to the most critical moving object can be counted, so as to obtain the first association number of the target detection box. Among them, for each frame of image whose MCDO observation target is associated with the target detection box, the first association number of the target detection box can be incremented by one.

[0132] Among them, to ensure the accuracy of the first association number, the number of times of continuously associated MCDO observation targets can be counted, that is, the first association number is the number of times the updated target detection box continuously associates with the camera observation target corresponding to the most critical moving object.

[0133] If the first association number is greater than the first threshold, it means that the state of the target detection box has been predicted and updated based on the relative motion parameters of the MCDO observation target at multiple moments, and the state of the target detection box can accurately represent the state of the MCDO observation target at this time. Therefore, it can be determined that the state of the target detection box converges at this time.

[0134] Alternatively, for any target detection box, the number of times of associating with the camera observation target can also be counted, and the convergence is judged based on this number. Specifically, step S205 "judging whether the state of the updated target detection box converges" can include steps B3 to B5.

[0135] Step B3, determining the second association number of the updated target detection box; the second association number is the number of times the updated target detection box associates with the camera observation target.

[0136] Step B4, when the second association number is greater than the second threshold, determining the covariance matrix between the relative speed and the relative depth corresponding to the updated target detection box.

[0137] Step B5, when the covariance matrix satisfies the convergence condition, determining that the state of the updated target detection box converges.

[0138] In this embodiment, for at least some of the target detection boxes, the number of times they associate with the camera observation target can be counted, and this number is used as the second association number of the target detection box. Among them, to avoid misjudgment, the second association number can specifically be the number of times the updated target detection box associates with the same camera observation target; for example, IDs can be set for different camera observation targets, and the second association number is the number of times the updated target detection box associates with the camera observation target corresponding to the same ID.

[0139] If the second association count is greater than the second threshold, it indicates that the target detection box has probably become stable. In addition, to ensure the accuracy of the convergence judgment, in this embodiment, the covariance matrix between the relative speed and relative depth corresponding to the target detection box is also determined, and the state convergence of the target detection box is further judged based on this covariance matrix.

[0140] Among them, the convergence condition of the covariance matrix is set in advance. If the second association count of the target detection box is greater than the second threshold and the covariance matrix meets the convergence condition, it can be determined that the state of the target detection box converges. For example, it can be judged whether to converge based on the rank of the covariance matrix or based on the residual.

[0141] For the target detection box associated with the MCDO observation target, the convergence judgment can also be performed again based on the above steps B3 to B5 to improve the reliability of the judgment result.

[0142] Optionally, the covariance matrix of the target detection box is:

[0143]

[0144] Among them, depth represents the relative depth observation, speed represents the relative speed observation, represents the covariance matrix between the relative speed and relative depth of P(depth, speed), and P dd represents the variance of the relative depth observation, and P ss represents the variance of the relative speed observation, and P ds represents the covariance between the relative depth observation and the relative speed observation, and P sd represents the covariance between the relative speed observation and the relative depth observation.

[0145] In this embodiment, for the current frame image collected by the camera, the relative depth and relative speed of the camera observation target in it can be determined, that is, the relative depth observation depth and the relative speed observation speed. For example, the current frame image can be input into the Velocity Net, and the relative depth and relative speed of the camera observation target can be determined based on the output of this velocity network. Based on the relative depth observation depth and the relative speed observation speed corresponding to the frame image, the covariance matrix between the relative speed and the relative depth can be calculated. Among them, P ds and P sd have different physical meanings, but their values are generally the same.

[0146] In this embodiment, the set convergence condition is specifically:

[0147]

[0148] Among them, depth′ is the relative depth of the target detection box, that is, the relative depth predicted by the target detection box. Both k1 and k2 are preset threshold parameters, and the two are hyperparameters set in advance.

[0149] After determining the covariance matrix of the target detection box, it is possible to simply and quickly determine whether the above formula holds: If P dd -P ds *P sd / P ss <(k1 * depth′) 2 , and P ss -P sd *P ds / P dd <k2 2 , then the covariance matrix satisfies the convergence condition, that is, the state of the target detection box converges.

[0150] In this embodiment, by counting the first association times of the target detection box, it is possible to quickly determine the convergence of the target detection box associated with the MCDO observation target, which is convenient for quickly determining whether the corresponding target detection box is stationary. By presetting the convergence condition of the covariance matrix, based on the second association times of the target detection box and the covariance matrix, it is possible to more accurately determine whether it converges, effectively reducing the possibility of misjudgment.

[0151] Optionally, based on the association times with the stationary radar observation target, it is also possible to judge the convergence of the corresponding target detection box. Specifically, step S205 "judging whether the state of the updated target detection box converges" may further include step B6.

[0152] Step B6, determining the third association times of the updated target detection box; the third association times is the number of times the updated target detection box is associated with the stationary radar observation target; in the case where the third association times is greater than the third threshold, it is determined that the state of the updated target detection box converges.

[0153] In this embodiment, similar to the process of judging the convergence of the target detection box based on the first association times above, after determining the third association times of the target detection box associated with the stationary radar observation target, it is possible to judge whether the third association times is greater than the third threshold. If the third association times is greater than the third threshold, it can be determined that the state of the updated target detection box converges.

[0154] Among them, the first threshold, the second threshold, and the third threshold may be the same or different, and can be specifically determined based on the actual situation.

[0155] Step S206, in the case where the state of the updated target detection box converges, determining whether the updated target detection box is stationary.

[0156] For details, please refer to Figure 1 step S104 of the embodiments shown, which will not be elaborated here.

[0157] In some alternative embodiments, the above step S206 "When the state of the updated target detection box converges, determine whether the updated target detection box is stationary" includes the following steps C1.

[0158] Step C1, when the state of the updated target detection box converges, if the relative speed of the updated target detection box is less than the current speed threshold, determine that the updated target detection box is stationary.

[0159] In this embodiment, if the observed target is stationary, the relative speed between it and itself should be the negative value of its own speed; for example, if the vehicle's own speed is 30 m / s, the relative speed between it and the stationary target should be -30 m / s. When the state of the target detection box converges, the smaller its relative speed is, the closer it is to the negative value of its own speed, and the more likely the target detection box is stationary.

[0160] Specifically, a corresponding speed threshold can be set based on the own speed, and this speed threshold can be slightly larger than the negative value of the own speed; if the relative speed of the target detection box is less than the current speed threshold, it can be considered that the relative speed of the target detection box is basically the same as the negative value of the own speed, and thus it can be simply and quickly determined that the target detection box is stationary.

[0161] For example, if it is determined based on steps B1 to B2 that the target detection box associated with the MCDO observed target converges, it can be continuously determined based on step C1 whether the target detection box associated with the MCDO observed target is stationary.

[0162] Alternatively, the above step S206 "When the state of the updated target detection box converges, determine whether the updated target detection box is stationary" can also include the following steps C2 to C3.

[0163] Step C2, when the state of the updated target detection box converges, with the negative value of the own speed as the variable and the relative speed of the updated target detection box as the expectation, determine the cumulative probability of the normal distribution corresponding to the own speed.

[0164] In this embodiment, for a target detection box with a converged state, the relative speed of the target detection box can be determined, and the own speed can also be determined. This own speed can be, for example, the speed of an autonomous vehicle. And, with the negative value of the own speed as the variable and the relative speed of the updated target detection box as the expectation, construct the probability density function of the cumulative probability of the normal distribution, so as to determine the cumulative probability of the normal distribution corresponding to this own speed.

[0165] Optionally, the cumulative probability of the normal distribution F(x; μ, σ) is as follows:

[0166]

[0167] where x represents the negative value of the self - speed, Δv represents the preset speed difference, μ represents the relative - speed expectation, and σ represents the relative - speed standard deviation.

[0168] In this embodiment, the speed difference Δv is preset, and the speed difference Δv can be dynamically adjusted, which can be specifically determined based on the actual situation. After detecting the self - speed v, the negative value of the self - speed v can be used as the variable x, that is, x = -v; and after determining the relative speed speed′ of the target detection frame, the relative speed speed′ can be used as the expectation μ, and the relative - speed standard deviation σ can be calculated based on the relative - speed observation values of multiple frames. It can be understood that the variance σ of the relative speed 2 , is also the variance P in the covariance matrix ss .

[0169] Step C3, when the cumulative probability of the normal distribution is greater than the preset probability value, determine that the updated target detection frame is stationary.

[0170] In this embodiment, a probability value for determining whether the target is stationary is preset, that is, the preset probability value, and the preset probability value can be specifically determined based on the relative - speed standard deviation σ. If the cumulative normal distribution F(x; μ, σ) corresponding to the negative value x of the self - speed is greater than the preset probability value (for example, 0.6 - 0.8), it can be explained that x is close to the relative speed μ, so at this time, it can be determined that the target detection frame is stationary.

[0171] Judging whether the target is stationary based on the cumulative probability of the normal distribution can improve the detection accuracy of stationary targets and avoid some moving targets being misjudged as stationary, thus causing abnormal deceleration of the autonomous vehicle.

[0172] Optionally, for the radar - observed target, it can be simply determined whether the associated target detection frame is stationary. Specifically, the above step "when the state of the updated target detection frame converges, determine whether the updated target detection frame is stationary" can include step C4.

[0173] Step C4, when the third association count of the updated target detection frame associated with the stationary radar - observed target is greater than the third threshold, determine that the updated target detection frame is stationary.

[0174] In this embodiment, if a certain target detection box is associated with a stationary radar observation target and the third association count of its associated stationary radar observation target is greater than the third threshold, since the radar can generally detect stationary targets more accurately, when the state of the target detection box converges, it can be directly determined that the target detection box is stationary.

[0175] Figure 4 A detailed flowchart of the detection method for the stationary target is shown, which combines a camera and a radar to detect the stationary target. As Figure 4 shown, the method includes the following steps.

[0176] Step S401, initialize the target detection box.

[0177] For example, create an initial target detection box for the observation target collected by the camera and perform Kalman filter state prediction to obtain the updated target detection box.

[0178] Step S402, extract the camera observation target.

[0179] For example, obtain the current frame image collected by the camera and identify the camera observation targets in the current frame image, specifically including MCDO observation targets and non-MCDO observation targets.

[0180] Step S403, the first round of association and update of MCDO observation targets.

[0181] Step S404, association and update of non-MCDO observation targets.

[0182] Step S405, create a target detection box for the remaining camera observation targets.

[0183] Step S406, the first round of association and update of unassociated MCDO observation targets.

[0184] Step S407, delete the invalid target detection boxes.

[0185] Among them, for the process of the above steps S403 to S407, specifically, refer to the relevant descriptions of the above steps A1 to A6, which will not be elaborated here.

[0186] Step S408, determine whether the first association count N1 is greater than the first threshold Th1. If N1 > Th1, it means that the state of the target detection box converges, and then execute the subsequent step S409.

[0187] Among them, after step S403, if the MCDO observation target is successfully associated, this step S408 can be executed.

[0188] Step S409, if the relative speed of the target detection box is less than the current speed threshold, then the target detection box is stationary. After that, execute the subsequent step S415.

[0189] Among them, for specific details, reference can be made to the relevant description in step C1, which will not be elaborated here.

[0190] In step S410, it is determined whether the second association count N2 is greater than the second threshold Th2. If N2 > Th2, the subsequent step S411 is continued.

[0191] In step S411, it is determined whether the covariance matrix P satisfies the convergence condition. If the covariance matrix P satisfies the convergence condition, it indicates that the state of the target detection box converges, and the subsequent step S412 is executed.

[0192] In step S412, the cumulative probability of the normal distribution corresponding to its own speed is determined. If the cumulative probability of the normal distribution is greater than the preset probability value, it is determined that the target detection box is stationary. Then, the subsequent step S415 is executed.

[0193] Among them, for specific details, reference can be made to the relevant description from step C2 to step C3, which will not be elaborated here.

[0194] In step S413, the stationary radar observation targets are extracted.

[0195] In step S414, it is determined whether the third association count N3 is greater than the third threshold Th3. If N3 > Th3, it is directly determined that the target detection box is stationary, and the subsequent step S415 is executed.

[0196] In step S415, the stationary target detection box is published.

[0197] For example, it is possible to send the target detection box of the stationary downstream abdomen to perform planning and control based on the stationary target detection box.

[0198] The method for detecting stationary targets provided by the embodiments of the present invention creates a target detection box using a camera, and combines the camera and radar to perform association update on the target detection box, determines whether the target is stationary, and has a high precision and recall rate for detecting stationary targets, which can effectively avoid the problem of missed detection of stationary vehicle targets caused by missed detection of long-distance cameras and missed detection of radar tracks.

[0199] In this embodiment, a device for detecting stationary targets is also provided. This device is used to implement the above embodiments and preferred implementation manners, and those that have been described will not be elaborated again. As used hereinafter, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0200] This embodiment provides a device for detecting stationary targets, as Figure 5 shown, including:

[0201] The target recognition module 501 is configured to obtain the current frame image collected by the camera and recognize the camera observation target in the current frame image;

[0202] The association and update module 502 is configured to associate the camera observation target with the current target detection box and perform state prediction update on the associated target detection box according to the relative motion parameters of the camera observation target; the target detection box includes the detection box created based on the observed historical target;

[0203] The convergence judgment module 503 is configured to judge whether the updated target detection box converges in state;

[0204] The detection module 504 is configured to determine whether the updated target detection box is stationary when the state of the updated target detection box converges.

[0205] In some alternative embodiments, the association and update module 502 associates the camera observation target with the current target detection box, including:

[0206] Associating the first camera observation target with the current target detection box; the first camera observation target is the camera observation target corresponding to the most critical moving object;

[0207] After that, associating the second camera observation target with the target detection box that is not currently associated with the first camera observation target; the second camera observation target is the other camera observation targets except the first camera observation target;

[0208] Creating a new target detection box for the second camera observation target that fails to associate with the target detection box.

[0209] In some alternative embodiments, the association and update module 502 associates the camera observation target with the current target detection box, further including:

[0210] In the case that the first camera observation target fails to associate with the target detection box, after creating a new target detection box for the second camera observation target that fails to associate with the target detection box, re-associating the first camera observation target with the target detection box that is not currently associated with the camera observation target;

[0211] In the case that the first camera observation target still fails to associate with the target detection box after re-association, creating a new target detection box for the first camera observation target.

[0212] In some alternative embodiments, the convergence judgment module 503 judges whether the updated target detection box converges in state, including:

[0213] Determine the first association count of the updated target detection box; the first association count is the number of times the updated target detection box is associated with the camera observation target corresponding to the most critical moving object;

[0214] When the first association count is greater than a first threshold, determine that the state of the updated target detection box converges;

[0215] And / or,

[0216] Determine the second association count of the updated target detection box; the second association count is the number of times the updated target detection box is associated with the camera observation target;

[0217] When the second association count is greater than a second threshold, determine the covariance matrix between the relative speed and relative depth corresponding to the updated target detection box;

[0218] When the covariance matrix satisfies the convergence condition, determine that the state of the updated target detection box converges.

[0219] In some alternative embodiments, the covariance matrix is:

[0220]

[0221] where depth represents the relative depth observation, speed represents the relative speed observation, represents the covariance matrix between the relative speed and relative depth of P(depth, speed), and P dd represents the variance of the relative depth observation, and P ss represents the variance of the relative speed observation, and P ds represents the covariance between the relative depth observation and the relative speed observation, and P sd represents the covariance between the relative speed observation and the relative depth observation;

[0222] The convergence condition includes:

[0223]

[0224] where depth′ is the relative depth of the target detection box, and both k1 and k2 are preset threshold parameters.

[0225] In some alternative embodiments, when the state of the updated target detection box converges, the detection module 504 determines whether the updated target detection box is stationary, including:

[0226] When the state of the updated target detection box converges, if the relative speed of the updated target detection box is less than the current speed threshold, determine that the updated target detection box is stationary;

[0227] Or,

[0228] When the state of the updated object detection box converges, taking the negative value of its own speed as a variable and the relative speed of the updated object detection box as an expectation, determine the cumulative probability of the normal distribution corresponding to the own speed;

[0229] When the cumulative probability of the normal distribution is greater than a preset probability value, determine that the updated object detection box is stationary.

[0230] In some alternative embodiments, the cumulative probability of the normal distribution is:

[0231]

[0232] Wherein, x represents the negative value of the own speed, Δv represents a preset speed difference, μ represents the relative speed expectation, and σ represents the relative speed standard deviation.

[0233] In some alternative embodiments, the target recognition module 501 is further configured to determine the radar observation target currently recognized by the radar and filter out the stationary radar observation targets;

[0234] The association update module 502 is further configured to associate the stationary radar observation targets with the current object detection boxes and perform state prediction update on the associated object detection boxes according to the relative motion parameters of the stationary radar observation targets.

[0235] In some alternative embodiments, when the convergence judgment module 503 judges whether the state of the updated object detection box converges, it includes:

[0236] Determine the third association count of the updated object detection box; the third association count is the number of times the updated object detection box associates with the stationary radar observation targets; when the third association count is greater than a third threshold, determine that the state of the updated object detection box converges;

[0237] When the state of the updated object detection box converges, the detection module 504 determines whether the updated object detection box is stationary, including:

[0238] When the third association count of the updated object detection box associating with the stationary radar observation targets is greater than the third threshold, determine that the updated object detection box is stationary.

[0239] In some alternative embodiments, the association update module 502 is further configured to:

[0240] Delete the object detection boxes that meet the failure conditions;

[0241] Among them, the failure conditions include: the duration for which the target detection box is not associated exceeds a first preset duration; or, the target type of the target detection box is inconsistent with the target type of the associated observed target, and the duration of the inconsistency exceeds a second preset duration.

[0242] The further function descriptions of the above-mentioned various modules and units are the same as those in the corresponding embodiments above, and will not be elaborated here.

[0243] The detection device for stationary targets in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, including a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0244] An embodiment of the present invention further provides a vehicle, which includes a vehicle controller that can implement the detection method for stationary targets provided in any of the above embodiments.

[0245] An embodiment of the present invention further provides a computer device having the above Figure 5 shown detection device for stationary targets. For example, the computer device can be a vehicle controller.

[0246] Please refer to Figure 6 , Figure 6 is a schematic structural diagram of a computer device provided by an optional embodiment of the present invention. As Figure 6 shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (such as an array of servers, a set of blade servers, or a multi-processor system). Figure 6 In

[0247] The processor 10 may be a central processing unit, a network processor, or a combination thereof. Among them, the processor 10 may further include a hardware chip. The above-mentioned hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The above-mentioned programmable logic device may be a complex programmable logic device, a field-programmable gate array, a generic array logic, or any combination thereof.

[0248] Among them, the memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiments.

[0249] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may further include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely disposed relative to the processor 10, and these remote memories may be connected to the computer device through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0250] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk, or a solid-state drive; the memory 20 may further include a combination of the above types of memories.

[0251] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0252] The embodiments of the present invention further provide a computer-readable storage medium. The method according to the embodiments of the present invention may be implemented in hardware, firmware, or may be implemented as computer code recorded on a storage medium, or may be implemented as computer code originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and to be stored in a local storage medium, so that the method described herein may be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium may be a magnetic disk, an optical disc, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium may further include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component capable of storing or receiving software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiments is implemented.

[0253] A part of the present invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the present invention through the operations of the computer. Those skilled in the art should understand that the forms in which computer program instructions exist in a computer-readable medium include but are not limited to source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include but are not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible by the computer.

[0254] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for detecting a stationary target, characterized in that: The method comprises: Acquire a current frame image captured by a camera, and identify a camera observation target in the current frame image; Associating the camera-observed target with a current target detection frame, and performing state prediction update on the associated target detection frame according to the relative motion parameters of the camera-observed target; the target detection frame includes a detection frame created based on a previously observed historical target; Determine whether the updated target detection frame is in a converged state based on the number of times the target detection frame is associated with the observed target; When the state of the updated target detection frame converges, it is determined whether the updated target detection frame is stationary.

2. The method according to claim 1, characterized in that The associating the camera observation target with the current target detection frame includes: Associating a first camera observation target with a current target detection frame; the first camera observation target is a camera observation target corresponding to the most critical moving object; Afterwards, associating the second camera observation target with the target detection frame that is not currently associated with the first camera observation target; the second camera observation target is a camera observation target other than the first camera observation target; A new object detection box is created for the second camera observation object that cannot be associated with the object detection box.

3. The method according to claim 2, characterized in that The associating the camera observation target with the current target detection frame also includes: In a case where the first camera observation target cannot be associated with a target detection frame, after creating a new target detection frame for the second camera observation target that cannot be associated with a target detection frame, re-associating the first camera observation target with a target detection frame that is not currently associated with the camera observation target; In a case where the first camera observation target cannot be associated with the target detection frame again, a new target detection frame is created for the first camera observation target.

4. The method according to claim 1, characterized in that The step of judging whether the updated target detection frame is in a converged state according to the number of times the target detection frame is associated with the observed target includes: Determine a first association number of the updated target detection frame; the first association number is the number of times the updated target detection frame is associated with the camera observation target corresponding to the most critical moving object; When the first association number is greater than a first threshold, determining that the updated target detection frame state converges; and / or, Determine a second association number of the updated target detection frame; the second association number is the number of times the updated target detection frame is associated with the camera observation target; When the second association number is greater than a second threshold, determining a covariance matrix between a relative speed and a relative depth corresponding to the updated target detection frame; When the covariance matrix satisfies a convergence condition, it is determined that the updated target detection frame state converges.

5. The method according to claim 4, characterized in that The covariance matrix is: Where depth represents the relative depth observation, speed represents the relative speed observation, and P(depth, speed) represents the covariance matrix between relative speed and relative depth. dd represents the variance of relative depth observations, P ss represents the variance of the relative velocity observation, P ds represents the covariance between the relative depth observation and the relative velocity observation, P sd represents the covariance between the relative velocity observation and the relative depth observation; The convergence conditions include: Among them, depth′ is the relative depth of the target detection box, and k1 and k2 are both preset threshold parameters.

6. The method according to claim 1, characterized in that When the state of the updated target detection frame converges, determining whether the updated target detection frame is stationary includes: When the state of the updated target detection frame converges, if the relative speed of the updated target detection frame is less than the current speed threshold, determining that the updated target detection frame is stationary; or, When the state of the updated target detection frame converges, the negative value of the own speed is used as a variable and the relative speed of the updated target detection frame is used as an expectation, and the normal distribution cumulative probability corresponding to the own speed is determined; When the normal distribution cumulative probability is greater than a preset probability value, it is determined that the target in the updated target detection frame is stationary.

7. The method according to claim 6, characterized in that The normal distribution cumulative probability is: Wherein, x represents the negative value of the own speed, Δv represents the preset speed difference, μ represents the relative speed expectation, and σ represents the relative speed standard deviation.

8. The method according to claim 1, characterized in that: Also includes: Determine the radar observation targets currently identified by the radar, and screen out the stationary radar observation targets; The stationary radar observation target is associated with the current target detection frame, and the state prediction and update of the associated target detection frame are performed according to the relative motion parameters of the stationary radar observation target.

9. The method according to claim 8, characterized in that The step of judging whether the updated target detection frame is in a converged state according to the number of times the target detection frame is associated with the observed target includes: Determine a third association number of the updated target detection frame; the third association number is the number of times the updated target detection frame is associated with the stationary radar observation target; when the third association number is greater than a third threshold, determine that the state of the updated target detection frame converges; When the state of the updated target detection frame converges, determining whether the updated target detection frame is stationary includes: When a third association number of the updated target detection frame and the stationary radar observation target is greater than a third threshold, it is determined that the updated target detection frame is stationary.

10. The method according to claim 1, characterized in that Also includes: Delete the target detection frame that meets the failure conditions; Among them, the failure conditions include: the time period during which the target detection frame is not associated exceeds a first preset time period; or the target type of the target detection frame is inconsistent with the target type of the associated observation target, and the inconsistency lasts for more than a second preset time period.

11. A stationary target detection device, characterized in that: The device comprises: The target recognition module is used to obtain the current frame image captured by the camera and identify the camera observation target in the current frame image; An association update module, used to associate the camera observation target with the current target detection frame, and perform state prediction update on the associated target detection frame according to the relative motion parameters of the camera observation target; the target detection frame includes a detection frame created based on a previously observed historical target; A convergence judgment module is used to judge whether the updated target detection frame is in a converged state according to the number of times the target detection frame is associated with the observed target; The detection module is used to determine whether the updated target detection frame is stationary when the state of the updated target detection frame converges.

12. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the stationary target detection method according to any one of claims 1 to 10 by executing the computer instructions.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the stationary target detection method according to any one of claims 1 to 10.

14. A vehicle, characterized in that: The vehicle comprises a vehicle controller, and the vehicle controller is used to execute the stationary target detection method according to any one of claims 1 to 10.

Images