False detection identification method, device and storage medium

Abstract

Embodiments of the present application provide a false detection identification method and device, and a storage medium, and relate to the field of intelligent driving and vehicle function safety monitoring. The method comprises: obtaining a first object list and a second object list, the first object list comprising candidate first objects determined through multi-sensor fusion processing, and the second object list comprising candidate second objects determined through a radar sensor; determining, from the first object list and the second object list, first objects and second objects located within a region of interest; matching a target object for a first object according to a spatial positional relationship between the first object and the second object, the target object being a second object with the shortest distance to the first object; and verifying the first object according to the target object to determine a verification result, the verification result being used to indicate whether the first object is a false detection object. The method is used to improve the robustness and determination certainty of false detection identification, reduce the risk of false triggering of automatic emergency braking, and improve braking decision reliability.

Description

Technical Field

[0001] This application relates to the field of intelligent driving and vehicle functional safety monitoring, and in particular to a method, device and storage medium for identifying false detections. Background Technology

[0002] Automatic emergency braking systems in intelligent driving typically rely on the perception fusion results of sensors such as cameras, lidar, and millimeter-wave radar to identify obstacles ahead and make contact risk assessments accordingly.

[0003] However, existing false detection protection methods mostly rely on the internal perception fusion or use attribute thresholds, global distance matching, etc., which are difficult to maintain independent effectiveness when the perception module is abnormal. At the same time, single space or attribute verification is easily affected by dense scenes, noise and accidental proximity, resulting in insufficient robustness of false detection identification, and the output results are usually relatively coarse.

[0004] Therefore, how to improve the independence, robustness, and quantifiability of false detection of perceived targets in automatic emergency braking scenarios has become an urgent technical problem to be solved. Summary of the Invention

[0005] This application provides a false detection identification method, device, and storage medium to improve the robustness and clarity of false detection identification, reduce the risk of false triggering of automatic emergency braking, and improve the reliability of braking decisions.

[0006] In a first aspect, embodiments of this application provide a method for identifying false detections, including:

[0007] Obtain a first object list and a second object list. The first object list includes candidate first objects determined by multi-sensor fusion processing, and the second object list includes candidate second objects determined by radar sensors.

[0008] From the first object list and the second object list, determine the first and second objects located within the region of interest;

[0009] Based on the spatial relationship between the first object and the second object, a target object is matched for the first object, and the target object is the second object that is closest to the first object;

[0010] The first object is verified based on the target object, and the verification result is determined. The verification result is used to indicate whether the first object is a false detection object.

[0011] In one possible implementation, matching the target object for the first object includes:

[0012] The expanded bounding box of the first object is determined based on the preset safety margin;

[0013] In the second object located within the expanded bounding box, determine the target object.

[0014] In one possible implementation, the method also includes:

[0015] Determine the displacement vector of the second object relative to the first object;

[0016] Based on the displacement vector and the heading angle of the first object, determine whether the second object is located within the expanded bounding box.

[0017] In one possible implementation, the method also includes:

[0018] When there is no second object in the expanded bounding box, determine the Euclidean distance between the first object and the second object;

[0019] The target object is determined based on the Euclidean distance and the distance threshold.

[0020] In one possible implementation, the first object is validated based on the target object, including:

[0021] If the target object exists, determine that the first object passes the existence verification.

[0022] In one possible implementation, the first object is validated based on the target object, including:

[0023] When the speed difference between the target object and the first object is less than the speed difference threshold, the first object is determined to have passed the speed verification.

[0024] In one possible implementation, the first object is validated based on the target object, including:

[0025] When the difference between the velocity direction angle of the target object and the heading angle of the first object is less than the direction threshold, the first object is determined to have passed the motion direction verification.

[0026] In one possible implementation, determining the verification result includes:

[0027] The results of existence verification, velocity verification, and / or direction of motion verification are converted into verification scores, respectively.

[0028] The verification result is determined based on the verification score and its weight.

[0029] Secondly, embodiments of this application provide a false detection identification device, comprising:

[0030] The acquisition module is used to acquire a first object list and a second object list. The first object list includes candidate first objects determined by multi-sensor fusion processing, and the second object list includes candidate second objects determined by radar sensors.

[0031] The first determining module is used to determine the first object and the second object located within the region of interest from the first object list and the second object list;

[0032] The processing module is used to match a target object for the first object based on the spatial relationship between the first object and the second object. The target object is the second object that is closest to the first object.

[0033] The second determination module is used to verify the first object based on the target object and determine the verification result. The verification result is used to indicate whether the first object is a false detection object.

[0034] In one possible implementation, the processing module is specifically used for:

[0035] The expanded bounding box of the first object is determined based on the preset safety margin;

[0036] In the second object located within the expanded bounding box, determine the target object.

[0037] In one possible implementation, the false detection identification device is also used for:

[0038] Determine the displacement vector of the second object relative to the first object;

[0039] Based on the displacement vector and the heading angle of the first object, determine whether the second object is located within the expanded bounding box.

[0040] In one possible implementation, the false detection identification device is also used for:

[0041] When there is no second object in the expanded bounding box, determine the Euclidean distance between the first object and the second object;

[0042] The target object is determined based on the Euclidean distance and the distance threshold.

[0043] In one possible implementation, the second determining module is specifically used for:

[0044] If the target object exists, determine that the first object passes the existence verification.

[0045] In one possible implementation, the second determining module is further used for:

[0046] When the speed difference between the target object and the first object is less than the speed difference threshold, the first object is determined to have passed the speed verification.

[0047] In one possible implementation, the second determining module is further used for:

[0048] When the difference between the velocity direction angle of the target object and the heading angle of the first object is less than the direction threshold, the first object is determined to have passed the motion direction verification.

[0049] In one possible implementation, the second determining module is further used for:

[0050] The results of existence verification, velocity verification, and / or direction of motion verification are converted into verification scores, respectively.

[0051] The verification result is determined based on the verification score and its weight.

[0052] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0053] The memory stores the instructions that the computer executes;

[0054] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0055] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0056] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0057] The false detection identification method, device, and storage medium provided in this application, by acquiring a first object list based on multi-sensor fusion processing and a second object list based on radar sensors, and determining the first and second objects located within the region of interest, can focus false detection identification on candidate targets within the automatic emergency braking area, reducing interference from irrelevant targets. By matching the first object with the shortest distance based on the spatial relationship between the first and second objects, and verifying the first object based on the target object, a verification result indicating whether the first object is a false detection object is determined. This allows the radar perception results to form an independent verification of the fused perception results, improving the robustness of false detection identification in dense scenes, noise interference, and perception anomalies, thereby improving the independence, accuracy, and quantifiability of false detection identification of perceived targets in automatic emergency braking scenarios. Attached Figure Description

[0058] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0059] Figure 1This application provides an illustration of an application scenario.

[0060] Figure 2 Flowchart of the false detection identification method provided in this application Figure 1 ;

[0061] Figure 3 Flowchart of the false detection identification method provided in this application Figure 2 ;

[0062] Figure 4 A schematic diagram illustrating the determination of a target object according to an embodiment of this application;

[0063] Figure 5 A schematic diagram of the false detection identification device provided in this application;

[0064] Figure 6 A schematic diagram of the structure of the electronic device provided in this application.

[0065] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0066] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0067] In the field of intelligent driving, automatic emergency braking typically relies on the perception fusion results of sensors such as cameras, LiDAR (Light Detection and Ranging), and millimeter-wave radar to identify obstacles in front of the vehicle and trigger a collision risk assessment. Such systems generally include a front-end perception fusion module, radar sensors, and a braking decision-making chain, and are applied to driving scenarios requiring continuous monitoring of targets ahead, such as urban following and highway cruising.

[0068] In existing technologies, false detection protection is typically based on the results of perception fusion. This involves determining the actual existence of a target through internal consistency checks, target attribute threshold checks, or matching perceived targets with radar targets globally by distance. The basic idea is to use existing perception outputs to perform secondary screening of candidate targets, thereby reducing the likelihood of erroneous braking.

[0069] However, most of the above methods rely on the internal operation of perception fusion. When the perception module itself experiences association anomalies, coordinate transformation anomalies, or false targets caused by environmental interference, the relevant verification mechanisms may also fail simultaneously, making it difficult to form independent and effective monitoring. At the same time, relying solely on a single attribute threshold or coarse distance matching is prone to erroneous associations in situations such as dense traffic, road surface reflections, and multipath echoes, causing non-existent targets to be mistakenly identified as verified, or making it impossible to accurately identify real targets due to unstable matching.

[0070] The direct consequence of this is that the automatic emergency braking system may generate unnecessary braking actions even when there are no real obstacles, affecting passenger comfort and potentially inducing rear-end collisions and other safety risks. Therefore, how to improve the reliability of false detection and the clarity of judgment criteria in automatic emergency braking scenarios, independently of the fused perception results, has become an urgent technical problem to be solved.

[0071] In view of this, this application provides a method for identifying false detections. By acquiring a first object list determined after multi-sensor fusion processing and a second object list determined by radar sensors, objects within the region of interest are first selected from the two types of objects. Then, based on the spatial relationship between the first object and the second object, the second object with the shortest distance is matched for the first object. The first object is then verified based on the target object to determine whether the first object is a false detection object.

[0072] In the application architecture of an automatic emergency braking system, this technical approach can be deployed after the front-end perception fusion module and radar sensors to independently verify the candidate targets output by the fusion. By introducing region of interest constraints, spatial positional matching, and target object-based verification mechanisms, the independence and robustness of false detection identification can be improved without relying on the internal verification logic of perception fusion, providing a more reliable input basis for subsequent braking decisions.

[0073] Figure 1 This is a schematic diagram illustrating an application scenario provided by this application. It includes a vehicle 1 and a falsely detected object 2. Vehicle 1 is a normally driving vehicle equipped with a front-end perception fusion module and a radar sensor. The front-end perception fusion module detects whether there are any objects around vehicle 1 that could affect normal driving, so as to trigger emergency braking if such an object is present. The radar sensor, such as millimeter-wave radar, operates independently of the front-end perception fusion module. Falsely detected object 2 is a spurious object falsely detected by the front-end perception fusion module due to road surface reflection or other reasons. If unnecessary emergency braking is triggered because the front-end perception fusion module detects a spurious object, it could lead to a rear-end collision or injury to occupants. In this case, the millimeter-wave radar, acting as an independent verification source, identifies falsely detected object 2 as a spurious object, thus avoiding triggering emergency braking.

[0074] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0075] Figure 2 Flowchart of the false detection identification method provided in this application Figure 1 .like Figure 2 As shown, the method includes:

[0076] S201. Obtain the first object list and the second object list.

[0077] The first object list includes candidate first objects determined through multi-sensor fusion processing, and the second object list includes candidate second objects determined by radar sensors.

[0078] The first object list is used to carry the candidate first objects determined by multi-sensor fusion processing in the front-end perception fusion module, serving as the input object set for subsequent screening, matching, and verification; the second object list is used to carry the candidate second objects determined by the radar sensor, serving as the reference object set for independent cross-verification of the first objects.

[0079] In this application, the executing entity can be a false detection identification control unit deployed in the automatic emergency braking system. This control unit is communicatively connected to the front-end perception fusion module and the radar processing module, and receives object data at fixed intervals during vehicle operation. For example, the AEB monitoring module's cycle is 100 milliseconds, or 10 Hz. Within each monitoring cycle, after completing the first object identification by AEB, this method is invoked to perform false detection identification on the first target.

[0080] In one possible embodiment, the first object list output by the front-end perception fusion module may include one or more candidate first objects, or it may be empty. An empty list indicates that the front-end perception fusion module has not detected an obstacle in front of the vehicle, and the subsequent false detection identification process is not triggered. For each candidate first object, at least the following are included: object identifier (id), position coordinates relative to the vehicle coordinate system (including longitudinal and lateral distances, in meters), and acceleration components (including lateral and longitudinal acceleration components, in meters). One or more of the following: heading angle (yaw, unit: radians), object category (e.g., enumeration values ​​of cars, trucks, pedestrians, etc.), and object confidence (value range: [0,1]).

[0081] The second object list output by the radar processing module may include one or more candidate second objects, or it may be empty. An empty list indicates that the radar processing module has not detected an obstacle in front of the vehicle. For each candidate second object, at least one of the following is included: position coordinates, position uncertainty (including lateral and longitudinal position uncertainty, in meters), velocity component, radar cross section (RCS, in dBsm), radar flag (valid, Boolean value), and bounding box dimensions (including length and width, in meters). The valid flag indicates whether the radar hardware has malfunctioned; when a malfunction occurs, the flag indicates that the first target is invalid, with a value of, for example, 0.

[0082] For example, the first object list can be received via vehicle Ethernet, CAN (Controller Area Network) bus, or on-chip shared memory, while the second object list can be received directly from the millimeter-wave radar target-level output interface or generated after parsing by the radar signal processor. The first candidate object is the object to be verified, representing candidate targets from the multi-sensor fusion results, and its spatial association with the radar-side object will be established subsequently; the second candidate object is the candidate target on the radar side, used to independently verify the first object.

[0083] Based on the above processing method, this step forms two sets of objects that are independent of each other and can be aligned in time, so that subsequent processing does not depend on the internal verification link of fusion, but directly uses the external correspondence between the fusion result and the radar result to carry out the review.

[0084] S202. From the first object list and the second object list, determine the first object and the second object located within the region of interest.

[0085] The Region of Interest (ROI) is used to limit the range of targets participating in false detection identification, avoiding calculations on objects in irrelevant regions. The first object is the candidate target on the fusion side that is retained and enters the verification process after being screened in this region, and the second object is the candidate target that participates in the verification as radar reference information after being screened in this region.

[0086] In this application, the false detection identification control unit pre-defines the spatial boundary of the region of interest in the vehicle coordinate system. The vehicle coordinate system uses the front axle center or the vehicle's center of gravity as its origin, with the longitudinal direction towards the vehicle's forward movement as positive and the lateral direction towards the left side of the vehicle as positive. Considering the use case of automatic emergency braking, the region of interest is set in front of the vehicle, and its boundary is defined by both the upper and lower limits of the longitudinal distance and the range of the lateral width. For example, in urban following scenarios, the longitudinal range can be set to 0 to 80 meters, and the lateral range can be set to 4 to 6 meters on each side of the lane centerline; in highway cruising scenarios, the longitudinal range can be extended to 0 to 150 meters, while the lateral range maintains a width corresponding to the coverage of the current lane and adjacent lanes.

[0087] For example, after determining the center coordinates C(cx,cy), length L, and width W (all in meters), the four boundary coordinates of the region of interest can be expressed as:

[0088]

[0089]

[0090]

[0091]

[0092] in, The minimum boundary point in the longitudinal direction. The point representing the maximum vertical boundary. The minimum horizontal boundary point, This is the maximum horizontal boundary point.

[0093] For example, given a center coordinate C(40,0), length L=60 meters, and width W=6 meters, the corresponding boundary coordinates are... =10 meters, =70 meters =-3 meters, =3 meters. This ROI covers an area 10 to 70 meters in front of the vehicle and 3 meters to the left and right, which is the frontal collision lane that the AEB system is most concerned about.

[0094] The method to determine whether a point P(px, py) is inside a rectangle ROI is: if px is greater than or equal to py, then px must be greater than or equal to py. And px is less than or equal to And py is greater than or equal to And py is less than or equal to Closed interval judgment (including equal signs) is used to ensure that no objects on the boundary are missed.

[0095] The process of identifying the first and second objects within the Region of Interest (ROI) involves: traversing the list of first objects, taking the center position P(x,y) of the Bird's Eye View (BEV) projection of each first object, and determining whether it lies within the ROI. If the point lies within the ROI, the candidate first object is identified as the first object. Similarly, the process of traversing the list of second objects skips candidate second objects with a false validity flag. For valid candidate second objects, the BEV position is taken, and the same determination is made regarding whether it lies within the ROI. If the point lies within the ROI, the candidate second object is identified as the second object. A false validity flag indicates a radar hardware failure; in this case, all first objects are marked as unverified (UNVALIDATED), and the process returns without further steps. This design ensures that no false judgments occur when the radar fails, adhering to the fail-safe principle.

[0096] Based on the above analysis, by first applying a unified region constraint to the first and second object lists, subsequent matching can be restricted to the same physical space of interest. This reduces the number of remote, lateral, or reverse irrelevant targets entering the matching process, lowers the probability of erroneous associations caused by global coarse matching, and ensures that subsequent verification results correspond to the target areas actually processed by automatic emergency braking. It should be understood that the above example is merely illustrative and not limiting.

[0097] S203. Based on the spatial relationship between the first object and the second object, match the first object with a target object, where the target object is the second object that is closest to the first object.

[0098] The target object represents the second object that is closest to the first object and selected for verifying the first object. It serves as a matching reference for the first object, supporting the determination of subsequent verification results. In this application, the false detection identification control unit sequentially traverses the second objects located within the region of interest for each first object within the region of interest and calculates the spatial distance between them. The spatial relationship can be determined by the difference in longitudinal and lateral coordinates between the two objects in the same vehicle coordinate system.

[0099] For example, Euclidean distance, weighted distance, etc. can be used as matching metrics. The weighted distance includes longitudinal deviation weight and lateral deviation weight. When automatic emergency braking pays more attention to longitudinal collision risk, the longitudinal deviation weight is set to be greater than the lateral deviation weight.

[0100] Based on the calculated spatial distance, the second object that is closest to the first object is selected from the second objects as the target object of the first object.

[0101] For example, the system can also maintain a cross-cycle matching cache. For a first object that has established a stable correspondence in the previous cycle, it prioritizes searching near its historical matching objects in the current cycle to reduce matching jumps caused by target intersections in dense traffic scenarios. The second object, as a candidate target on the radar side, serves to provide existence and motion information support independent of the fused perception results; the target object is the specific correspondence of this supporting information on a single first object.

[0102] In one possible implementation, Mahalanobis distance can be used to match the first object with the target object. A covariance matrix is ​​constructed using the positional uncertainty of the second object, and the Mahalanobis distance between the first and second objects is calculated. A successful match is determined when the Mahalanobis distance is less than a threshold (e.g., 3 standard deviations). This method can adaptively adjust the matching tolerance based on radar measurement accuracy, resulting in more uniform matching effects at different distances and angles.

[0103] The above matching method matches each first target individually. Another possible implementation is to use a globally optimal matching strategy to match target objects for each first object. First, the distance matrix between all first targets and all second targets is calculated. Then, the Hungarian Algorithm is used to solve for the minimum weight bipartite graph matching, obtaining the globally optimal one-to-one association result. This method avoids the local optima problem caused by object-by-object greedy matching (such as two first objects competing for the same second object), and may result in higher association quality in dense scenarios.

[0104] Based on the above analysis, this step establishes the shortest distance matching relationship between the first object and the second object within the region of interest, and maps the candidate targets on the fusion side to the independent observation results on the radar side one-to-one, no longer relying on the internal correlation results of the fusion, so that the subsequent verification is based on the externally measurable spatial relationship.

[0105] S204. Verify the first object based on the target object, and determine the verification result. The verification result is used to indicate whether the first object is a false detection object.

[0106] The verification results are used to transform the verification process of the target object against the first object into a final judgment output, and to provide a basis for downstream automatic emergency braking decisions. In this application, for each first object that has been matched with the target object, the false detection identification control unit performs a review process based on the correspondence between the target object and the first object to determine whether the first object is a false detection object.

[0107] For example, the association information between the first object and the target object can be compared, and the verification result can be determined based on the comparison result. When the target object can provide effective support for the first object, the verification result indicates that the first object is not a false detection object; when the target object cannot provide effective support for the first object, the verification result indicates that the first object is a false detection object. For cases where there is an anomaly in a single period but stable matching between preceding and following periods, a time smoothing mechanism can be introduced to merge the current period result with historical results, preventing instantaneous missed detections or short-term echo fluctuations from causing the verification result to jitter.

[0108] Based on the above analysis, verifying the first object against the target object essentially involves externally confirming the authenticity of the fused candidate target using independent radar observations. When the target object can maintain a correspondence with the first object, it indicates that the first object has radar support and should not be identified as a false detection. When the target object cannot provide effective support for the first object, it indicates that the first object lacks independent radar evidence and can be identified as a false detection. This process transforms the criteria for false detection judgment from simply relying on the internal output of the fusion to a clear criterion based on independent cross-sensor verification, thereby improving the reliability of false detection identification in automatic emergency braking scenarios.

[0109] This application provides a method for identifying false detections, including obtaining a first object list and a second object list. The first object list includes candidate first objects determined through multi-sensor fusion processing, and the second object list includes candidate second objects determined by a radar sensor. From the first and second object lists, a first object and a second object located within a region of interest are determined. Based on the spatial relationship between the first and second objects, a target object is matched for the first object, where the target object is the second object with the shortest distance to the first object. The first object is then verified based on the target object, and a verification result is determined, which indicates whether the first object is a false detection. In this application, by performing unified time alignment, region of interest filtering, nearest neighbor matching based on spatial relationship, and independent verification based on the target object on both the fusion-side and radar-side objects, an external verification link is formed after the front-end perception fusion module, making false detection identification independent of the internal verification mechanism of the fusion module. For false fused targets caused by correlation anomalies, coordinate transformation anomalies, environmental reflections, or multipath echoes, this method can output false detection judgments when there is a lack of corresponding radar target support or insufficient correspondence. For real targets ahead, it can provide verification support through the matched target objects, thereby providing directly usable verification results for subsequent decisions on automatic emergency braking.

[0110] Figure 3 Flowchart of the false detection identification method provided in this application Figure 2 ,like Figure 3 As shown, the method includes:

[0111] S301. Get the first object list and the second object list.

[0112] S302. From the first object list and the second object list, determine the first object and the second object located within the region of interest.

[0113] For determining the target object from the first and second objects within a region of interest, this application provides a dual-method spatial association strategy that prioritizes bounding box inclusion determination and uses Euclidean distance matching as a fallback. First, bounding box inclusion determination is performed in steps S303 to S306. If the bounding box inclusion determination determines that a second object meeting the conditions exists, these second objects are identified as strong matching candidates, and the target object is determined from among these strong matching candidates. If the bounding box inclusion determination determines that no second object meeting the conditions exists, Euclidean distance matching is further performed as shown in step S307. The second objects identified by this matching method are identified as weak matching candidates, and the target object is determined from among these weak matching candidates. These two strategies are described in detail below.

[0114] S303. Determine the extended bounding box of the first object according to the preset safety margin.

[0115] Determining the extended bounding box of the first object includes: determining the initial bounding box of the first object, adjusting the initial bounding box according to the safety margin, and determining the extended bounding box.

[0116] The initial bounding box is determined by the information carried by each first object in the first object list. For a first object O, its initial bounding box consists of the center coordinates (ox, oy), length Lo, width Wo, and heading angle. definition.

[0117] The safety margin is preset to compensate for radar ranging accuracy errors and initial bounding box estimation errors, preventing the exclusion of radar points that should be matched due to minor measurement deviations. The safety margin is denoted as m, for example, 0.5 meters. The safety margin is added to the length and width of the initial bounding box to obtain the expanded bounding box.

[0118] Extended bounding box half length The halfwidth can be represented as:

[0119]

[0120]

[0121] S304. Determine the displacement vector of the second object relative to the first object.

[0122] Let the second object be R(rx, ry). Then the displacement vector of the second object relative to the center coordinates of the bounding box of the first object can be expressed as:

[0123]

[0124]

[0125] Where dx is the longitudinal displacement vector and dy is the lateral displacement vector.

[0126] S305. Based on the displacement vector and the heading angle of the first object, determine whether the second object is located within the extended bounding box.

[0127] First, the displacement vector is transformed into the local coordinate system of the first target through a reverse rotation. The local coordinates can be represented as:

[0128]

[0129]

[0130] in, For local ordinates, This represents the local x-axis.

[0131] The physical meaning of this transformation is: to rotate the coordinates of the second object from the vehicle coordinate system to a local coordinate system with the orientation of the first object itself as the positive X-axis, so that the bounding box is an axis-aligned rectangle in the local coordinate system, thereby simplifying the inclusion determination calculation.

[0132] Determining whether local coordinates are within the bounding box includes: if and If so, it is determined that the second object is located within the expanded bounding box of the first object.

[0133] When the second object is located within the expanded bounding box, i.e., there is at least one second object within the expanded bounding box, proceed to step S306.

[0134] If no second object exists in the expanded bounding box, proceed to step S307.

[0135] In the above method, by combining the displacement vector and the heading angle of the first object to perform the expanded bounding box determination, the screening result of the second object can simultaneously reflect the relative distance and orientation relationship, thereby making the target matching more consistent with the actual spatial attitude of the first object and providing a more stable geometric basis for subsequent verification.

[0136] S306. In the second object located within the expanded bounding box, determine the target object.

[0137] The target object is the second object that is closest to the center coordinates of the first object among all the second objects in the expanded bounding box.

[0138] S307. Determine the Euclidean distance between the first object and the second object, and determine the target object based on the Euclidean distance and the distance threshold.

[0139] The Euclidean distance d can be expressed as:

[0140]

[0141] If d is less than the distance threshold If so, the second object is identified as a weak matching candidate. The distance threshold is, for example, 2 meters.

[0142] When multiple weak matching candidates exist, the second object that is closest to the first object is identified as the target object.

[0143] The Euclidean distance fallback method is used to ensure that matching can still be completed in the following situations: (a) the initial bounding box size of small objects (such as pedestrians and bicycles) is small and may not cover the radar scattering point even with a safety margin; (b) the radar scattering point is located on the surface of the object rather than near the center, which is off the range of the initial bounding box; (c) there is a large error in the initial bounding box size or heading angle estimation of the first object.

[0144] Since Euclidean distance directly reflects the spatial proximity between targets, and the distance threshold constrains the acceptable deviation, this supplementary matching method can maintain the consistency of object association even when the extended bounding box is not hit, thus providing a stable source of input for the verification results.

[0145] When this method is adopted, if the local geometric relationship results in the expansion bounding box not containing a second object, the system can still complete the object association through Euclidean distance and determine the target object or output the result of no target object based on this. This makes the judgment criteria for false detection clearer and is independent of the internal verification logic of the fusion module.

[0146] See Figure 4 , Figure 4 This is a schematic diagram illustrating target object determination provided in an embodiment of this application. In the vehicle coordinate system, with the rear axle center as the origin, the X-axis points to the front of the vehicle, and the Y-axis points to the left side of the vehicle. The four boundaries of the region of interest are x: [10m, 70m], y: [-3m, 3m], as shown by the outermost dashed box in the figure. O represents the first object, and the center coordinates of its initial bounding box are o(ox, oy), as shown by the solid box in the figure. After adding a safety margin of 0.5m to the initial bounding box, the expanded bounding box is shown by the dashed box surrounding the initial bounding box.

[0147] Within the region of interest, a first object O, a second object R1, a second object R2, and another second object R3 coexist. If R1 is located inside the expanded bounding box, then R1 is identified as a strong matching candidate and confirmed as the target object.

[0148] If the first target is not present in the expanded bounding box (i.e., R1 is absent), the system further determines whether a weak matching candidate exists. Taking a distance threshold of 2 meters, the Euclidean distance d2 between R2 and the first target is 1.8 meters, which is less than the distance threshold, making it a weak matching candidate. The Euclidean distance d3 between R3 and the first target is 4.5 meters, which is greater than the distance threshold, making it not a weak matching candidate. Therefore, R2 is identified as the target object.

[0149] If no target object is matched for the first target after traversing all the second targets, it is determined that the first target has no radar support.

[0150] S308. Verify the first object based on the target object.

[0151] Optionally, the first object can be verified, which may include three items: existence verification, velocity verification, and motion direction verification. These three verifications can be performed sequentially.

[0152] For example, existence verification includes: when a target object exists, determining that a first object passes existence verification.

[0153] Existence verification is the first line of defense against false detections, verifying whether the first object has been independently confirmed by the radar sensor. The judgment logic is as follows: if a target object is found for the first object in step S305, i.e., there is radar support, then the existence verification passes, indicating that the radar sensor has also detected the target, and the probability of false detection is greatly reduced; if no target object is found, i.e., there is no radar support, then the existence verification fails, indicating that the first object has no corresponding echo in the radar field of view, which is highly suspected to be a false detection. The physical basis is: millimeter-wave radar, as an ASIL-B level independent sensor, uses a completely different physical principle (electromagnetic wave reflection) from LiDAR and cameras, and is not affected by conditions such as light, rain, and fog. If a target is only detected by LiDAR / camera but there is no echo from millimeter-wave radar, it is highly likely to be a false detection caused by factors such as road surface reflection, shadows, or multipath echoes.

[0154] This method uses the presence or absence of a corresponding target object as the verification criterion, allowing the verification result to directly reflect whether the first object possesses radar support, thus forming an independent confirmation of the first object. With this method, the verification result of the first object no longer relies solely on internal consistency of the fusion process, but is directly provided by the existence of the target object. This makes the criteria for false detection identification clearer and provides more stable input results for subsequent braking decisions.

[0155] For example, speed verification includes: determining that the first object passes speed verification when the speed difference between the target object and the first object is less than a speed difference threshold.

[0156] If the existence verification fails, the speed verification is immediately deemed a failure. If the existence verification passes, the speed verification is then performed.

[0157] Speed ​​verification includes: determining the speed scalar of the first object; determining the speed scalar of the target object; determining the speed difference; determining that the first object passes speed verification when the speed difference is less than the speed threshold, and determining that the first object fails speed verification when the speed difference is greater than or equal to the speed threshold.

[0158] Specifically, the velocity scalar vf of the first object can be expressed as:

[0159]

[0160] in, Let be the velocity component of the first object in the longitudinal direction. Let be the velocity component of the first object in the transverse direction.

[0161] The velocity scalar vr of the target object can be expressed as:

[0162]

[0163] in, This represents the longitudinal velocity component of the target object. This represents the velocity component of the target object in the horizontal direction.

[0164] Speed ​​difference It can be represented as:

[0165]

[0166] The speed threshold is denoted as The velocity threshold can be determined based on the accuracy of the millimeter-wave radar and the front-end perception fusion module. For example, the velocity threshold is 3 m / s. The value is based on the fact that the typical velocity measurement accuracy of the millimeter-wave radar is ±0.5 m / s (directly measured by the Doppler effect), and the velocity estimation accuracy of the front-end perception fusion module is about ±1 m / s (obtained by continuous frame object position difference and Kalman filtering). The maximum reasonable deviation between the two is about ±1.5 m / s, which is 3 m / s after taking a safety factor of 2.

[0167] That is when If the speed verification is successful, it will pass; otherwise, it will fail.

[0168] Using the above method, the verification of the first object is not only based on spatial positional relationship, but also constrained by velocity consistency. This ensures that the matched target object and the first object are consistent in motion characteristics before the pass result is output, thereby improving the reliability of the first object verification result and providing a more stable basis for subsequent false detection object judgment.

[0169] For example, motion direction verification includes: determining that the first object passes motion direction verification when the difference between the velocity angle direction of the target object and the heading angle of the first object is less than a direction threshold.

[0170] If the existence verification fails, the motion direction verification is also automatically rejected. If the existence verification passes, the motion direction verification is then performed.

[0171] Furthermore, when the target object's speed is less than the minimum speed threshold, the motion direction verification is directly deemed successful. The minimum speed threshold is, for example, 0.5 m / s. Since the noise in the velocity direction is extremely high at low speeds, the motion direction verification is passed by default.

[0172] When the speed of the target object is greater than or equal to the minimum speed threshold, the motion direction verification includes: determining the speed direction angle of the target object; determining the heading angle of the first object; determining the heading angle difference; if the heading angle difference is less than the direction threshold, determining that the first object has passed the motion direction verification; otherwise, determining that the first object has failed the motion direction verification.

[0173] Specifically, the velocity direction angle of the target object It can be represented as:

[0174]

[0175] in, It is the arctangent function in the four quadrants.

[0176] The heading angle of the first object It can be represented as:

[0177]

[0178] Heading angle difference It can be represented as:

[0179]

[0180] Here, norm is a function that normalizes the angle to the range [-π, π].

[0181] Since the first object may be in a forward or reverse state, the velocity direction of the target object may be the same as the velocity direction of the first object (forward, with a deviation of less than 0 degrees) or opposite (reverse, with a deviation of about 180 degrees). Therefore, it is necessary to check both forward and reverse deviation at the same time.

[0182] Positive deviation can be expressed as:

[0183]

[0184] The reverse bias can be expressed as:

[0185]

[0186] Take the smaller value between the positive and negative deviations and the direction threshold. If a comparison is made, If the first object passes the motion direction verification, then the first object is determined to have passed the motion direction verification; otherwise, the first object is determined to have failed the motion direction verification.

[0187] Direction threshold For example, 0.785 radians, approximately 45 degrees, takes into account the following factors: the velocity azimuth accuracy of millimeter-wave radar is affected by radar installation angle error, multipath effect, and object shape, with a typical azimuth accuracy of approximately ±10 to ±20 degrees; the heading accuracy of the front-end perception fusion module is approximately ±5 to ±10 degrees; the maximum reasonable deviation between the two is approximately ±30 degrees, which, after taking a safety factor of 1.5, is approximately 45 degrees.

[0188] In another possible implementation, motion direction verification includes: determining the cosine similarity between the velocity vector of the first target and the velocity vector of the target object; when the absolute value of the cosine similarity is greater than the similarity threshold, the motion direction verification is successful.

[0189] Cosine similarity can be expressed as:

[0190]

[0191] The similarity threshold is, for example, 0.707, corresponding to the cosine of a 45-degree angle. This method can utilize the velocity vector information from two sensor sources simultaneously, rather than relying solely on the heading angle estimation of the first object, which is advantageous in scenarios where the heading angle estimation is inaccurate but the velocity vector estimation is accurate.

[0192] By comparing the velocity direction of the target object with the heading angle of the first object, the verification result of the first object is simultaneously constrained by both spatial matching and motion trend relationships, thus making the verification conclusion closer to the target's actual motion state. This method reduces misjudgments caused by relying solely on positional proximity, making the final output verification result of the first object more suitable as a basis for target confirmation in automatic emergency braking scenarios.

[0193] S309. Convert the results of existence verification, velocity verification and / or motion direction verification into verification scores respectively.

[0194] Each type of verification corresponds to a verification score. For example, the verification score is 1 when the existence verification is passed and 0 when it fails; the verification score is 1 when the speed verification is passed and 0 when it fails; the verification score is 1 when the direction of motion verification is passed and 0 when it fails.

[0195] S310. Determine the verification result based on the verification score and score weight.

[0196] Determining the validation results includes: determining the confidence score based on the validation score, determining the validation result based on the confidence score, and the validation result includes suspected false positives and non-false positives.

[0197] The confidence score S can be expressed as:

[0198]

[0199] in, The weights for existence verification. Weights for speed verification Weights for verifying the direction of motion. The verification score is the score for existence verification. The verification score is for speed verification. This is the verification score for verifying the direction of motion. The sum of all weights is 1, therefore the confidence score S ranges from 0 to 1. The score for radar support markers is 1 when there is radar support, i.e., a target object exists; and 0 when there is no radar support, i.e., no target object exists.

[0200] Each weight can be preset according to the importance of each validation item. For example It is 0.5. It is 0.3. The weight is set to 0.2. The basis for the weight setting is that existence verification is the most basic verification (without radar support, subsequent verifications are meaningless), so it is given the highest weight of 0.5; velocity verification has higher measurement accuracy and more direct safety correlation than motion direction verification (AEB decision directly depends on the relative velocity of the object), so it is given the second highest weight of 0.3; motion direction verification is used as an auxiliary supplementary verification, so it is given the lowest weight of 0.2.

[0201] When there is no radar support , , the confidence scoring formula degenerates to S=0, which means completely untrustworthy; when there is radar support but the speed or direction check fails, S is an intermediate value between 0.5 and 0.8; when all three checks are passed, S=1, which means completely trustworthy.

[0202] The final verification result adopts the logical AND determination. When the existence verification, speed verification and movement direction verification are all passed, it indicates that the first object is confirmed as a non-false detection through radar cross-verification, AEB can safely rely on this target for braking decision-making, and further update the first target to the VERIFIED state.

[0203] Otherwise, the first object is determined as a suspected false detection, and its state is updated to SUSPICIOUS. The specific meanings are: (a) completely without radar support (S=0), the first object has no corresponding echo in the radar field of view, which is a highly suspected false detection, and the braking decision-making of AEB based on this object shall be suppressed; (b) there is radar support but the motion attributes are inconsistent (0<S<1), although the radar detects an echo nearby, the speed or direction does not match that of the first object. It may be that the front-end perception fusion module incorrectly associates an unrelated radar echo with the false detection object, which still requires vigilance.

[0204] For the first object that does not enter the ROI, the UNVALIDATED state is maintained, and it does not participate in the false detection identification and determination of this method.

[0205] For each first object in the ROI, there is a corresponding verification result respectively. The verification result can include the following information: object identification number, object position, verification status (one of verified, suspicious, unvalidated), whether there is a radar support flag, position and speed information of the target object, matching distance, verification scores of each of the three verifications, and confidence score.

[0206] In this embodiment, when the existence verification is passed and both the speed verification and movement direction verification meet the preset conditions, the system can generate a relatively high comprehensive verification result; when only part of the verification items are passed, the system can reduce the comprehensive verification result according to the corresponding weights; when none of the verification items are passed, a lower comprehensive verification result is output. Through the above weighting method, the verification result can comprehensively reflect the matching degree of the target object in terms of existence verification, speed consistency and movement direction consistency, and serve as a basis for determining whether the first object is a false detection.

[0207] After adopting this method, the verification result no longer depends on a single verification item, but forms a comprehensive judgment through the combination of verification scores and score weights. This enables the role of different verification dimensions in the determination to be clear and controllable, thereby improving the discrimination and stability of the verification result, and providing a clear numerical basis for subsequent false detection identification.

[0208] Optionally, the method of this application can further summarize information, including the total number of candidate first objects, the total number of first objects in the ROI, the number of verified first objects, and the number of suspicious first objects.

[0209] Furthermore, the verification results and summary information are input into the downstream AEB risk aggregation module to provide it with sufficient false detection identification information. In the AEB monitoring system of this embodiment, after performing false detection identification on the first AEB object using this method, corresponding risk signals are generated based on the verification status and confidence score: the first object in the VERIFIED state is confirmed as a non-false detection and is accepted for subsequent Time To Collision (TTC) calculation and braking decision; the first target in the SUSPICIOUS state is marked as a suspected false detection, triggering a braking suppression risk flag, and entering the weighted risk aggregation (with a weight of 0.12) to participate in the final risk level determination. When this flag is true, the braking decision of the AEB system will be suppressed or downgraded to avoid improper emergency braking due to false detection.

[0210] The following is a specific example illustrating the false detection identification method provided in this application.

[0211] Scenario conditions: The car is traveling at 50 km / h (approximately 13.9 m / s) on a city road. 30 meters ahead, a sedan is traveling in the same direction at 40 km / h (approximately 11.1 m / s) with a heading angle of 0 degrees (in the same direction as the car).

[0212] The first object list includes one candidate first object with the following basic information: object identifier id=42, ordinate x=30 meters, abscissa y=-0.5 meters, and longitudinal velocity component. =11.1 m / s, lateral velocity component =0 m / s, initial bounding box length L=4.5 m, width W=1.8 m, yaw=0 radians.

[0213] The list of second objects includes one candidate second object with the following basic information: vertical coordinate x = 30.3 meters, horizontal coordinate y = -0.2 meters, and vertical velocity component. =11.5 m / s, lateral velocity component =0.1 m / s, radar flag valid=true.

[0214] The center of the region of interest is (40, 0), with a length of 60 meters and a width of 6 meters.

[0215] First, determine whether the candidate first and second targets are located within the ROI: the vertical boundary of the ROI is [10, 70], and the horizontal boundary is [-3, 3]. The candidate first object (30, -0.5) is within the ROI, so it is retained and identified as the first object. The candidate second object (30.3, -0.2) is within the ROI, so it is retained and identified as the second object.

[0216] Perform a bounding box inclusion check on the second object (30.3, -0.2). The initial bounding box center of the first object is (30, -0.5), with a length of 4.5 meters, a width of 1.8 meters, a heading angle of 0 radians, and a safety margin of 0.5 meters. The extended bounding box half-length = (4.5 + 1) / 2 = 2.75 meters, and half-width = (1.8 + 1) / 2 = 1.4 meters. The displacement vectors dx = 0.3 and dy = 0.3. Since the heading angle is 0, the local coordinates... =0.3, =0.3. Judgment: |0.3|=0.3 is less than 2.75 and |0.3|=0.3 is less than 1.4, the second object is within the expanded bounding box, matching is successful, and the matching distance is... rice.

[0217] Existence verification: Supported by radar, passed, verification score is 1.

[0218] Speed ​​verification: meters per second meters per second m / s, less than the speed threshold of 3 m / s, passed, verification score is 1.

[0219] Motion direction verification: The target object's velocity scalar vr = 11.5 m / s, which is greater than the minimum velocity threshold of 0.5 m / s, so perform direction verification. Radius (approximately 0.5 degrees). radian. , , The radians are less than the direction threshold of 0.785 radians, so the verification score is 1.

[0220] The confidence score S = 0.5 × 1 + 1 × (0.3 × 1 + 0.2 × 1) = 1. Since all three verifications pass, the first object is not a false positive, and the status of the first object is updated to VERIFIED.

[0221] The first object, confirmed by radar cross-validation as a real, non-false detection (a vehicle actually in front), allows the AEB system to safely perform subsequent TTC calculations and braking decisions based on the object's attributes. If the front-end perception fusion module generates a false detection object due to road surface reflection (e.g., id=99, x=25 meters, y=0 meters, but no second object matches it), this false detection object will be classified as SUSPICIOUS (S=0, highly suspected false detection). The AEB system will then suppress braking decisions for this target, thus avoiding inappropriate emergency braking caused by false detection.

[0222] Figure 5 A schematic diagram of the false detection identification device provided in this application is shown below. Figure 5 As shown, the false detection identification device 50 provided in this embodiment includes:

[0223] The acquisition module 51 is used to acquire a first object list and a second object list. The first object list includes candidate first objects determined by multi-sensor fusion processing, and the second object list includes candidate second objects determined by radar sensors.

[0224] The first determining module 52 is used to determine the first object and the second object located in the region of interest from the first object list and the second object list;

[0225] Processing module 53 is used to match a target object for the first object based on the spatial relationship between the first object and the second object, wherein the target object is the second object that is closest to the first object;

[0226] The second determining module 54 is used to verify the first object based on the target object and determine the verification result. The verification result is used to indicate whether the first object is a false detection object.

[0227] In one possible implementation, processing module 53 is specifically used for:

[0228] The expanded bounding box of the first object is determined based on the preset safety margin;

[0229] In the second object located within the expanded bounding box, determine the target object.

[0230] In one possible implementation, the false detection identification device 50 is further used for:

[0231] Determine the displacement vector of the second object relative to the first object;

[0232] Based on the displacement vector and the heading angle of the first object, determine whether the second object is located within the expanded bounding box.

[0233] In one possible implementation, the false detection identification device 50 is further used for:

[0234] When there is no second object in the expanded bounding box, determine the Euclidean distance between the first object and the second object;

[0235] The target object is determined based on the Euclidean distance and the distance threshold.

[0236] In one possible implementation, the second determining module 54 is specifically used for:

[0237] If the target object exists, determine that the first object passes the existence verification.

[0238] In one possible implementation, the second determining module is further used for:

[0239] When the speed difference between the target object and the first object is less than the speed difference threshold, the first object is determined to have passed the speed verification.

[0240] In one possible implementation, the second determining module 54 is further configured to:

[0241] When the difference between the velocity direction angle of the target object and the heading angle of the first object is less than the direction threshold, the first object is determined to have passed the motion direction verification.

[0242] In one possible implementation, the second determining module 54 is further configured to:

[0243] The results of existence verification, velocity verification, and / or direction of motion verification are converted into verification scores, respectively.

[0244] The verification result is determined based on the verification score and its weight.

[0245] The false detection identification device provided in this embodiment can execute the false detection identification method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0246] Figure 6 A schematic diagram of the structure of the electronic device provided in this application. Figure 6 As shown, the electronic device 60 provided in this embodiment includes at least one processor 61 and a memory 62. Optionally, the device 60 further includes a communication component 63. The processor 61, memory 62, and communication component 63 are connected via a bus.

[0247] In the specific implementation process, at least one processor 61 executes computer execution instructions stored in memory 62, causing at least one processor 61 to execute the above-described false detection identification method.

[0248] The specific implementation process of processor 61 can be found in the above-described false detection identification method embodiment, which has a similar implementation principle and technical effect, and will not be repeated here.

[0249] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0250] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0251] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0252] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described false detection identification method.

[0253] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the aforementioned false detection identification method.

[0254] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0255] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an application-specific integrated circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0256] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0257] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0258] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0259] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0260] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0261] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for identifying false detections, characterized in that, include: Obtain a first object list and a second object list. The first object list includes candidate first objects determined by multi-sensor fusion processing, and the second object list includes candidate second objects determined by radar sensors. From the first object list and the second object list, determine the first object and the second object located within the region of interest; Based on the spatial relationship between the first object and the second object, a target object is matched for the first object, wherein the target object is the second object that is closest to the first object; The first object is verified based on the target object, and the verification result is determined. The verification result is used to indicate whether the first object is a false detection object.

2. The method according to claim 1, characterized in that, The step of matching the first object with a target object includes: The expanded bounding box of the first object is determined based on a preset safety margin; The target object is determined within the second object located in the expanded bounding box.

3. The method according to claim 2, characterized in that, The method further includes: Determine the displacement vector of the second object relative to the first object; Based on the displacement vector and the heading angle of the first object, determine whether the second object is located within the extended bounding box.

4. The method according to claim 2, characterized in that, The method further includes: When there is no second object in the extended bounding box, determine the Euclidean distance between the first object and the second object; The target object is determined based on the Euclidean distance and the distance threshold.

5. The method according to claim 1, characterized in that, The step of verifying the first object based on the target object includes: If the target object exists, the first object is determined to have passed the existence verification.

6. The method according to claim 1, characterized in that, The step of verifying the first object based on the target object includes: When the speed difference between the target object and the first object is less than the speed difference threshold, the first object is determined to have passed the speed verification.

7. The method according to claim 1, characterized in that, The step of verifying the first object based on the target object includes: When the difference between the velocity direction angle of the target object and the heading angle of the first object is less than a direction threshold, the first object is determined to have passed the motion direction verification.

8. The method according to any one of claims 1-7, characterized in that, The determination of the verification result includes: The results of existence verification, velocity verification, and / or direction of motion verification are converted into verification scores, respectively. The verification result is determined based on the verification score and the score weight.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-8.

Images