Target detection methods, apparatus, computer equipment and storage media

Abstract

This application relates to a target detection method, apparatus, computer device, and storage medium. The method includes: acquiring a point cloud dataset obtained by detecting the driving environment of a target vehicle; the point cloud dataset contains multiple frames of point cloud data; determining point cloud data within each object detection region based on the location information of each point cloud data in the point cloud dataset; selecting target point cloud data that meets the point cloud data feature conditions of the target object from the point cloud data within each object detection region based on the feature information of the point cloud data within each object detection region; determining the object detection confidence level corresponding to each object detection region based on the number of target point cloud data within each object detection region; and determining the target region where the target object is located based on the object detection region where the object detection confidence level is greater than or equal to a preset confidence threshold. This method can detect target objects before they emerge from an occluded area, improving driving safety.

Description

Technical Field

[0001] This application relates to the field of radar technology, and in particular to a target detection method, apparatus, computer equipment, and storage medium. Background Technology

[0002] The "ghost peek" scenario refers to a situation where pedestrians, vehicles, or other targets suddenly emerge from an obstructed area in front of the driver's vehicle (such as behind a parked vehicle on the roadside or behind a pillar in a garage). In such situations, drivers are prone to collisions due to insufficient time to avoid them.

[0003] In related technologies, vehicle-mounted radar can be used to perceive the driving environment, detecting targets such as pedestrians and vehicles. Drivers or autonomous vehicles can then adjust their driving strategies in a timely manner based on the detection results to avoid collisions. However, radar sensors typically have difficulty detecting targets that suddenly cross from obstructed areas. They usually only reliably detect targets after they have already moved a certain distance beyond the obstructed area. This means it's difficult to detect a target before it completely exits the obstructed area, potentially leading to a collision due to insufficient time to adjust the driving strategy. Summary of the Invention

[0004] Therefore, it is necessary to provide a target detection method, apparatus, computer device, and computer-readable storage medium that can detect a target object before it passes through an obstructed area, thereby improving driving safety, in order to address the aforementioned technical problems.

[0005] Firstly, this application provides a target detection method. The method includes:

[0006] A point cloud dataset is obtained by detecting the driving environment of the target vehicle; the point cloud dataset contains multiple frames of point cloud data.

[0007] Based on the location information of each point cloud data in the point cloud dataset, determine the point cloud data within the detection area of ​​each object;

[0008] Based on the feature information of the point cloud data in each object detection area, target point cloud data that meets the point cloud data feature conditions of the target object is selected from the point cloud data in each object detection area.

[0009] Based on the number of target point cloud data within each object detection region, the object detection confidence level corresponding to each object detection region is determined, and based on the object detection regions whose object detection confidence level is greater than or equal to a preset confidence threshold, the target region where the target object is located is determined.

[0010] In one embodiment, determining the point cloud data within each object detection area based on the location information of each point cloud data in the point cloud dataset includes:

[0011] The target detection area of ​​the driving environment is divided into grids to obtain multiple object detection areas;

[0012] The location information of each point cloud data in the point cloud dataset is matched with the location information of each object detection region to determine the point cloud data within each object detection region.

[0013] In one embodiment, determining the object detection confidence level corresponding to each object detection region based on the number of target point cloud data within each object detection region, and determining the target region where the target object is located based on the object detection regions whose object detection confidence level is greater than or equal to a preset confidence threshold, includes:

[0014] For each object detection region, the number of target point cloud data within the object detection region is determined as the object detection confidence of the object detection region, and a confidence threshold corresponding to the object detection region is obtained; the confidence threshold corresponding to the object detection region is negatively correlated with the distance between the object detection region and the target vehicle;

[0015] If the object detection confidence level in the object detection region is greater than or equal to the confidence threshold corresponding to the object detection region, the object detection region is determined as the target region where the target object is located.

[0016] In one embodiment, obtaining the confidence threshold corresponding to the object detection region includes:

[0017] Based on the distance between the object detection area and the target vehicle, the confidence threshold corresponding to the object detection area is determined in a pre-established correspondence between distance and confidence threshold.

[0018] In one embodiment, obtaining the confidence threshold corresponding to the object detection region includes:

[0019] Based on the location information of the object detection area, the grid area corresponding to the object detection area is determined;

[0020] In the pre-established correspondence between grid regions and confidence thresholds, the confidence threshold corresponding to the object detection region is obtained.

[0021] In one embodiment, determining the target region where the target object is located based on the object detection region whose object detection confidence is greater than or equal to a preset confidence threshold includes:

[0022] The object detection region with an object detection confidence level greater than or equal to a preset confidence threshold is determined as the first candidate region where the target object is located;

[0023] For each first candidate region, in the object detection regions adjacent to the first candidate region and the first candidate region, a region containing a number of target point cloud data that meets a preset quantity condition is determined as the target region where the target object is located.

[0024] In one embodiment, determining the target region where the target object is located based on the object detection region whose object detection confidence is greater than or equal to a preset confidence threshold includes:

[0025] The object detection region with an object detection confidence score greater than or equal to a preset confidence threshold is determined as the second candidate region where the target object is located;

[0026] In each of the second candidate regions, a pair of second candidate regions that meet the similarity condition between the multipath target and the real target is determined based on the feature information of the target point cloud data in each of the second candidate regions.

[0027] For the second candidate region pair, the second candidate region containing the most target point cloud data is taken as the target region where the target object is located.

[0028] In one embodiment, the method further includes:

[0029] If there are no other second candidate regions in each of the second candidate regions that satisfy the similarity condition between the multipath target and the real target, then the second candidate region of the target is taken as the target region where the target object is located.

[0030] Secondly, this application also provides a target detection device. The device includes:

[0031] The acquisition module is used to acquire a point cloud dataset obtained by detecting the driving environment of the target vehicle; the point cloud dataset contains multiple frames of point cloud data.

[0032] The first determining module is used to determine the point cloud data within the detection area of ​​each object based on the location information of each point cloud data in the point cloud dataset.

[0033] The filtering module is used to filter out target point cloud data that meets the point cloud data feature conditions of the target object from the point cloud data in each object detection area based on the feature information of the point cloud data in each object detection area.

[0034] The second determining module is used to determine the object detection confidence level corresponding to each object detection region based on the number of target point cloud data in each object detection region, and to determine the target region where the target object is located based on the object detection regions whose object detection confidence level is greater than or equal to a preset confidence threshold.

[0035] In one embodiment, the first determining module is specifically used for:

[0036] The target detection area of ​​the driving environment is divided into grids to obtain multiple object detection areas; the position information of each point cloud data in the point cloud dataset is matched with the position information of each object detection area to determine the point cloud data within each object detection area.

[0037] In one embodiment, the second determining module is specifically used for:

[0038] For each object detection region, the number of target point cloud data within the object detection region is determined as the object detection confidence of the object detection region, and a confidence threshold corresponding to the object detection region is obtained; the confidence threshold corresponding to the object detection region is negatively correlated with the distance between the object detection region and the target vehicle; if the object detection confidence of the object detection region is greater than or equal to the confidence threshold corresponding to the object detection region, the object detection region is determined as the target region where the target object is located.

[0039] In one embodiment, the second determining module is specifically used to: determine the confidence threshold corresponding to the object detection area based on the distance between the object detection area and the target vehicle, in a pre-established correspondence between distance and confidence threshold.

[0040] In one embodiment, the second determining module is specifically used to: determine the grid region corresponding to the object detection region based on the location information of the object detection region; and obtain the confidence threshold corresponding to the object detection region from a pre-established correspondence between grid regions and confidence thresholds.

[0041] In one embodiment, the second determining module is specifically used for:

[0042] The object detection region with an object detection confidence level greater than or equal to a preset confidence threshold is determined as the first candidate region where the target object is located; for each first candidate region, the region containing the number of target point cloud data that meets the preset quantity condition is determined among the object detection regions adjacent to the first candidate region and the first candidate region, and is taken as the target region where the target object is located.

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

[0044] The object detection region with an object detection confidence level greater than or equal to a preset confidence threshold is determined as the second candidate region where the target object is located. In each second candidate region, a pair of second candidate regions that meet the similarity condition between the multipath target and the real target is determined based on the feature information of the target point cloud data in each second candidate region. For the second candidate region pair, the second candidate region containing the most target point cloud data is taken as the target region where the target object is located.

[0045] In one embodiment, the apparatus further includes a third determining module, configured to, if in each of the second candidate regions there are no other second candidate regions that satisfy the similarity condition between the multipath target and the real target with the target second candidate region, then take the target second candidate region as the target region where the target object is located.

[0046] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described in the first aspect.

[0047] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0048] The aforementioned target detection method, apparatus, computer equipment, and storage medium acquire a point cloud dataset by detecting the driving environment of the target vehicle. Then, based on the location information of each point cloud data point in the dataset, the point cloud data within each object detection area is determined. Next, based on the feature information of the point cloud data, target point cloud data that meets the feature conditions of the target object is selected. Finally, based on the number of target point cloud data points within each object detection area, the object detection confidence level corresponding to each object detection area is determined. Finally, based on object detection areas where the object detection confidence level is greater than or equal to a preset confidence threshold, the target area where the target object is located is determined. The point cloud dataset contains multiple frames of point cloud data.

[0049] In this scheme, accumulating point cloud data from multiple frames increases the quantity and density of point cloud data for detecting "ghost peek" targets, preventing missed detections due to sparse point clouds in a single frame caused by occlusion. Furthermore, in ghost peek scenarios, the point cloud dataset can include multipath reflection point cloud data of the target object. By setting feature conditions for the target object's point cloud data, target point cloud data is filtered out to fully utilize multipath echoes reflected from the ground, walls, and other static objects in the surrounding environment to detect occluded targets. In addition, the object detection confidence is determined based on the quantity of target point cloud data within the object detection area. The detection area that meets the confidence criteria is considered the area where the target object is located, ensuring the accuracy of target detection. Therefore, this scheme can detect the location of the target object before it exits the occlusion area, allowing for earlier detection of ghost peek targets and providing drivers with more time to adjust their driving strategies, thus improving driving safety. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of a ghost peeking out from behind an obstacle, as shown in the example.

[0051] Figure 2 This is a flowchart illustrating a target detection method in one embodiment;

[0052] Figure 3 This is a flowchart illustrating the process of determining point cloud data within the detection area of ​​each object in one embodiment;

[0053] Figure 4 This is a schematic diagram of multiple grid regions in an example;

[0054] Figure 5 This is a flowchart illustrating the process of determining the target area where the target object is located in one embodiment;

[0055] Figure 6 This is a schematic diagram of the first candidate region and its adjacent regions in an example.

[0056] Figure 7 This is a flowchart illustrating the process of determining the target area where the target object is located in another embodiment;

[0057] Figure 8 This is a schematic diagram of the second candidate region in an example;

[0058] Figure 9 This is a structural block diagram of a target detection device in one embodiment;

[0059] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0061] First, before introducing the technical solutions of the embodiments of this application in detail, let's first introduce the technical background or technical evolution on which the embodiments of this application are based. Ghost peek scenario, such as... Figure 1 As shown, this refers to situations where pedestrians, vehicles, or other targets suddenly emerge from an obstructed area in front of a driving vehicle (such as behind a parked car or behind a pillar in a garage). In such cases, the driving vehicle is prone to collisions due to insufficient time to avoid them. Related technologies utilize vehicle-mounted radar to perceive the driving environment and detect pedestrians, vehicles, and other targets. The driver or autonomous vehicle can then adjust its driving strategy accordingly to avoid collisions. However, in "ghost peek" scenarios, the target is crossing horizontally or diagonally, resulting in a lower radial velocity relative to the radar. Furthermore, the target emerges from an obstructed area during its crossing, reducing the number of detectable points. This makes it difficult for the radar to detect targets suddenly appearing from an obstructed area (referred to as "ghost peek targets") in a timely manner. Generally, the target can only be reliably detected after it has already crossed a certain distance from the obstructed area. In other words, it is difficult to detect the target before it completely exits the obstructed area, potentially leading to insufficient time to adjust the driving strategy and resulting in a collision. Against this backdrop, through long-term research and development and experimental verification, the applicant has proposed the target detection method of this application. This method can detect the location of a target before it emerges from an obstructed area, allowing the driver more time to adjust their driving strategy and improving driving safety. Furthermore, it should be noted that the applicant has devoted considerable creative effort to discovering the technical problem of this application and to developing the technical solutions described in the following embodiments.

[0062] The target detection method provided in this application can be applied to terminals, including vehicle-mounted terminals or other electronic devices. For example, it can be a vehicle-mounted radar such as millimeter-wave radar, other vehicle-mounted terminals communicating with the vehicle-mounted radar, or electronic devices such as security radar or traffic radar. The terminal can acquire point cloud data detected by the radar of the driving environment, and then detect whether there are cross-traversing target objects (ghosting targets) in the driving environment, and if so, the target area where the target object is located. Therefore, the terminal can output the detection results to remind the driver to drive safely, or use the detection results to formulate driving strategies for the intelligent driving system to improve driving safety.

[0063] In one embodiment, such as Figure 2 As shown, a target detection method is provided. Taking the application of this method to a terminal as an example, the method includes the following steps:

[0064] Step 201: Obtain the point cloud dataset obtained by detecting the driving environment of the target vehicle.

[0065] In implementation, during the target vehicle's movement, onboard radar (such as 4D forward-facing millimeter-wave radar) can detect the vehicle's driving environment at a preset frame period to obtain point cloud data corresponding to various objects or pedestrians in the driving environment. Point cloud data can include location information (such as 3D coordinates), point cloud velocity, point cloud height, point cloud intensity, and point cloud azimuth. The point cloud dataset contains multiple frames of point cloud data, which can be consecutive frames or non-consecutive frames. Specifically, the terminal can acquire the point cloud data detected in the current frame (referred to as the current point cloud data) and the point cloud data detected in a preset number of historical frames preceding the current frame (referred to as historical point cloud data) for further processing. In some examples, the point cloud dataset contains 6 to 10 frames of point cloud data.

[0066] In "ghost peek" scenarios, the target object (the ghost peek target) is often obscured by objects in the environment, such as parked vehicles or garage pillars. This means the target object's environment is complex. The radar may receive echoes directly reflected from the target object, as well as multipath echoes jointly reflected by the target object and other static objects in the environment, such as the ground, pillars, and ceiling (in a basement scene). The point cloud data corresponding to these multipath echoes can be called multipath reflection point cloud data. Therefore, the point cloud dataset can contain the multipath reflection point cloud data of the target object.

[0067] Step 202: Determine the point cloud data within the detection area of ​​each object based on the location information of each point cloud data in the point cloud dataset.

[0068] The object detection area is a unit area within the target vehicle's driving environment used to detect target objects. For example, a pre-defined target detection area in the driving environment can be divided into multiple unit areas, each of which is the object detection area, to detect the presence of a target object within each unit area. The size of each unit area can be the same or different; for example, the size of each unit area can be positively or negatively correlated with the distance between that area and the target vehicle. The target detection area is generally the area along the target vehicle's driving path, where "ghosting" targets may appear, affecting the target vehicle's safe driving; therefore, target detection is necessary. The specific detection range can be set as needed.

[0069] In practice, the terminal can determine the point cloud data within the detection area of ​​each object based on the location information of each point cloud data in the point cloud dataset, that is, determine the point cloud data whose location information falls within the object detection area.

[0070] Understandably, since the positional information of the current point cloud data (point cloud data detected in the current frame) and the positional information of the object detection area are generally represented by their relative position to the target vehicle at the current moment, while the positional information of the historical point cloud data (point cloud data detected in historical frames) is generally represented by their relative position to the target vehicle at a historical moment (the detection moment corresponding to the historical frame), it is necessary to unify the coordinate system of each positional information. For example, the positional information of the historical point cloud data can be compensated based on the vehicle's attitude information, speed information, etc., at a historical moment, so as to map the positional information of each historical point cloud data to the coordinate system used in the current frame. In one example, the positional information of the historical point cloud data detected in the previous frame can be denoted as (Rx0, Ry0, Rz0), where Rx0, Ry0, and Rz0 represent the lateral distance, longitudinal distance, and height difference between the historical point cloud and the target vehicle at the previous moment (the moment corresponding to the previous frame), respectively. The time difference between the previous moment and the current moment is the radar frame period T. Assuming the target vehicle is in a state of uniform acceleration, the positional information of the historical point cloud data can be compensated based on the target vehicle's speed at the previous moment (denoted as V0), its speed at the current moment (denoted as V1), and the radar frame period T. The longitudinal distance (denoted as Ry1) of the target relative to the target vehicle at the current moment, corresponding to the historical point cloud data, can then be obtained as: Ry1 = Ry0 + 0.5 * (V0 + V1) * T. Therefore, the positional information of each point cloud data point and the positional information of the object detection area in the point cloud dataset can be unified into a single coordinate system. This allows the point cloud data within each object detection area to be determined based on the positional information after coordinate system unification.

[0071] Step 203: Based on the feature information of the point cloud data within each object detection area, select the target point cloud data that meets the feature conditions of the target object from the point cloud data within each object detection area.

[0072] The feature information of point cloud data can include point cloud height, point cloud intensity, point cloud velocity, etc. Correspondingly, the feature conditions of the target object's point cloud data can include one or more of the following: point cloud height conditions (such as point cloud height range), point cloud intensity conditions (such as point cloud intensity range), and point cloud velocity conditions (such as point cloud velocity range).

[0073] In implementation, the terminal can filter the point cloud data within each object detection area based on the feature information of the point cloud data within that area and the preset point cloud data feature conditions for the target object, thus selecting the target point cloud data within each object detection area. Specifically, the point cloud data feature conditions for the target object can be one or more of the following: point cloud height condition, point cloud intensity condition, and point cloud velocity condition. For example, if the target object has only one point cloud data feature condition, such as a point cloud height condition, the point cloud height of each point cloud data can be compared or matched with the point cloud height condition, and the point cloud data whose point cloud height meets the point cloud height condition can be determined as the target point cloud data. If the point cloud data of the target object has multiple feature conditions (such as point cloud height, point cloud intensity, and point cloud velocity), then the point cloud height, point cloud intensity, and point cloud velocity of each point cloud data point can be compared or matched with the point cloud height condition, point cloud intensity condition, and point cloud velocity condition, respectively. If the point cloud data meets the point cloud height condition, the point cloud intensity condition, and the point cloud velocity condition, then the point cloud data can be identified as the target point cloud data. The point cloud data filtering conditions for each object detection region can be the same, or filtering conditions can be set separately for each object detection region.

[0074] The point cloud data feature conditions for the target object can be pre-set based on experiments or experience. For example, point cloud data with and without "ghost peek" targets can be collected and analyzed to determine the point cloud height, velocity, and intensity characteristics in both scenarios, in order to set point cloud data feature conditions applicable to "ghost peek" scenarios. For instance, if in a scenario with "ghost peek" targets, 90% of the point cloud height is between -2m and 0.5m, then the point cloud height condition can be set to -2m to 0.5m (the point cloud height of the target point cloud data must be within this height range).

[0075] The settings for point cloud height, velocity, and intensity conditions can affect the accuracy and sensitivity of target detection. In some examples, the point cloud height condition can be set to -4m to 4m, -3m to 3m, -2m to 2m, -2m to 0m, or 2 to 3m (any one of these two ranges is acceptable). Point cloud data with negative height may be point cloud data detected by radar through ground reflection (multipath reflection point cloud data); point cloud data with positive height and greater than 2m may be point cloud data detected by radar through reflection from objects such as ceilings (multipath reflection point cloud data).

[0076] When detecting peeking targets, it relies more on point cloud data with lower intensity. Therefore, the point cloud intensity condition can be set to a lower intensity than that in conventional radar applications. For example, when detecting pedestrians in conventional radar applications, the point cloud intensity condition is generally set to greater than or equal to 50dB. In peeking scenarios, it can be set to 20dB to 30dB (0.4 to 0.6 times that of conventional scenarios). That is, the intensity of the target point cloud data should be between 20dB and 30dB.

[0077] For ghost-peek target detection, since most targets are in a horizontal or diagonal crossing motion (mostly horizontal), the crossing speed of pedestrians or vehicles is generally low. Unlike traditional target detection algorithms that need to consider the effectiveness of low, medium, and high-speed point clouds simultaneously, this solution can select only the low-speed point cloud data while still distinguishing stationary objects. For example, the point cloud velocity condition can be set to -3m / s to -0.2m / s or 0.2m / s to 3m / s (any one of these two ranges is sufficient), or it can be set to -3m / s to -0.4m / s or 0.4m / s to 3m / s, or even -1m / s to -0.4m / s or 0.4m / s to 1m / s. Here, point cloud velocity refers to the velocity relative to the ground.

[0078] Step 204: Determine the object detection confidence level corresponding to each object detection area based on the number of target point cloud data within each object detection area, and determine the target area where the target object is located based on the object detection areas where the object detection confidence level is greater than or equal to the preset confidence threshold.

[0079] In implementation, after filtering the target point cloud data within each object detection area, the terminal can determine the object detection confidence level corresponding to each object detection area based on the quantity of target point cloud data within that area. The object detection confidence level represents the likelihood that a target object exists within that detection area. For example, the quantity of target point cloud data can be directly used as the object detection confidence level, or the quantity of target point cloud data can be normalized to obtain the object detection confidence level. If the object detection confidence level corresponding to an object detection area is greater than or equal to a preset confidence threshold, then that object detection area can be identified as the target area where the target object is located, i.e., the presence of a target object (ghosting target) is considered to exist in that detection area. Since the point cloud data detected by radar may contain clutter noise, a confidence threshold can be set to filter each object detection area based on the object detection confidence level to ensure detection accuracy.

[0080] In the aforementioned target detection method, accumulating point cloud data from multiple frames increases the quantity and density of point cloud data for detecting "ghost peek" targets, preventing missed detections due to occlusion of the target and resulting sparse point cloud in a single frame. Furthermore, in ghost peek scenarios, the point cloud dataset can include multipath reflection point cloud data of the target object. By setting feature conditions for the target object's point cloud data, target point cloud data can be filtered out to fully utilize multipath echoes reflected from the ground, walls, and other static objects in the surrounding environment to detect occluded targets. In addition, the object detection confidence is determined based on the quantity of target point cloud data within the object detection area. The detection area that meets the confidence criteria is considered the area where the target object is located, ensuring the accuracy of target detection. Therefore, this scheme can detect the location of the target object before it exits the occlusion area, enabling relatively early detection of ghost peek targets and allowing drivers more time to adjust their driving strategies, thus improving driving safety.

[0081] In one embodiment, such as Figure 3 As shown, the process of determining the point cloud data within the detection area of ​​each object in step 202 specifically includes the following steps:

[0082] Step 301: Divide the target detection area of ​​the driving environment into a grid to obtain multiple object detection areas.

[0083] The target detection area for the driving environment refers to the radar detection area used to detect targets in the driving environment of the target vehicle at any given moment. The specific detection range can be set according to the actual situation. For example, if the target vehicle is traveling forward on the road, a rectangular area 60 meters long and 16 meters wide (8 meters each on the left and right sides of the target vehicle) in front of the target vehicle can be used as the target detection area to detect potential "ghosting" targets that may appear on the target vehicle's path.

[0084] In practice, the terminal can divide the target detection area into grids, resulting in multiple grid regions, such as... Figure 4 The diagram shows a grid area. The rules for dividing the grid area can be set according to the situation. For example, each grid area can be set to 2 meters by 2 meters. For the target detection area in the previous example (a rectangular area 60 meters long and 16 meters wide), this 60-meter by 16-meter target detection area can be divided into 240 grid areas. The location information of each grid area can be represented by the lateral distance Rx and the longitudinal distance Ry between the area and the target vehicle, such as... Figure 4The position information of the first grid cell in the first row from bottom to top can be represented as follows: the horizontal distance range is Rx ≥ -8m and Rx ≤ -6m, and the vertical distance range is Ry ≥ 0m and Ry ≤ 2m. Each grid area can be used as an object detection area to detect whether a target object exists in each object detection area.

[0085] Step 302: Match the location information of each point cloud data in the point cloud dataset with the location information of each object detection area to determine the point cloud data within each object detection area.

[0086] In implementation, the terminal can determine the point cloud data within each object detection region based on the location information of each point cloud data in the current point cloud dataset, the location information of each point cloud data in the historical point cloud dataset, and the location information of the object detection region. The coordinate system for each location information needs to be unified; generally, the location information can be represented by the lateral and longitudinal distances relative to the target vehicle at the current moment. For example, if the lateral distance range between object detection region i and the target vehicle is Rx ≤ 4m and Rx ≥ 2m, and the longitudinal distance range is Ry ≤ 4m and Ry ≥ 2m, and the lateral distance of point cloud data j relative to the target vehicle is Rx = 2.5m and the longitudinal distance is Ry = 3m, then point cloud data j is the point cloud data within object detection region i.

[0087] In this embodiment, the target detection area is divided into multiple grid regions by dividing the target detection area into grid regions, which serve as object detection areas. This allows for the accumulation of point clouds in each grid region across multiple frames. The presence of a target object in a grid region is determined based on the amount of point cloud data in that grid region. This enables the detection of the grid region where the target is located before it passes through the occlusion area, giving the vehicle more time to adjust its driving strategy and improving driving safety.

[0088] In one embodiment, the process of determining the target region where the target object is located in step 204 specifically includes the following steps: for each object detection region, the number of target point cloud data in the object detection region is determined as the object detection confidence of the object detection region, and the confidence threshold corresponding to the object detection region is obtained; if the object detection confidence of the object detection region is greater than or equal to the confidence threshold corresponding to the object detection region, the object detection region is determined as the target region where the target object is located.

[0089] Among them, the confidence threshold corresponding to the object detection area is negatively correlated with the distance between the object detection area and the target vehicle.

[0090] In implementation, for each object detection region, the terminal can determine the object detection confidence level of that region by the number of target point cloud data points within that region. Furthermore, the terminal can obtain the confidence threshold for each object detection region. This confidence threshold is negatively correlated with the distance between the object detection region and the target vehicle; that is, the closer the object detection region is to the target vehicle, the higher the confidence threshold, and vice versa. If the object detection confidence level of an object detection region is greater than or equal to the corresponding confidence threshold, the terminal can determine that object detection region as the target region containing the target object.

[0091] In one implementation, a correspondence between distance and confidence threshold can be established in advance. For example, the correspondence between distance and confidence threshold can be stored as a correspondence table between distance and confidence threshold. Then, the terminal can query the corresponding confidence threshold in the correspondence table based on the distance between the object detection area and the target vehicle, thereby determining the confidence threshold corresponding to the object detection area.

[0092] In one example, the correspondence between distance and confidence threshold is shown in Table 1. The longitudinal distance and lateral distance represent the lateral and longitudinal distances between the object detection area and the target vehicle, respectively. The confidence threshold represents the limit on the number of target point cloud data within the object detection area. For example, for an object detection area with location information {Rx≤4m, and Rx≥2m, Ry≤4m, and Ry≥2m}, according to Table 1, the confidence threshold corresponding to this object detection area is 5. If the number of target point cloud data within this object detection area is 6, that is, the object detection confidence corresponding to this object detection area is 6. At this time, the object detection confidence (6) of this object detection area is greater than or equal to the confidence threshold (5), so it can be considered that this object detection area is the area where the target object is located, that is, it can be considered that a ghost target exists within this object detection area. If the number of target point cloud data within this object detection area is 2, it can be considered that no ghost target exists within this object detection area.

[0093] Table 1. Correspondence between distance and confidence threshold

[0094]

[0095] In another implementation, the terminal can determine the corresponding grid region based on the location information of the object detection area. For example, the radar detection range (e.g., a 60m x 16m area) can be pre-divided into multiple grid regions (e.g., 240 grid regions). The object detection area can then be mapped to a grid region based on the location information of both the object detection area and the grid regions. Next, a correspondence between the grid regions and confidence thresholds can be established, such as by creating a table mapping grid identifiers to confidence thresholds. This mapping can be obtained from the distance-confidence threshold correspondence shown in Table 1. Therefore, the terminal can determine the corresponding grid region based on the object detection area's location information and directly obtain the corresponding confidence threshold from the pre-established mapping between grid regions and confidence thresholds.

[0096] In this embodiment, because the echo from a "ghost peek" target is relatively strong in areas close to the target vehicle, a higher point threshold (confidence threshold) is set to avoid excessive clutter at close range leading to false detections, thereby improving target detection accuracy. Conversely, for areas at greater distances, where the echo from a "ghost peek" target is relatively weak, a lower point threshold is set to ensure sufficient detection sensitivity. This balances both target detection accuracy and detection sensitivity.

[0097] In one embodiment, such as Figure 5 As shown, the process of determining the target area where the target object is located in step 204 specifically includes the following steps:

[0098] Step 501: The object detection region with an object detection confidence level greater than or equal to the preset confidence threshold is determined as the first candidate region where the target object is located.

[0099] In implementation, the terminal can compare the object detection confidence of each object detection region with a preset confidence threshold. If the object detection confidence of an object detection region is greater than or equal to the preset confidence threshold, the terminal can determine the object detection region as the first candidate region where the target object is located for subsequent processing; if the object detection confidence of an object detection region is less than the preset confidence threshold, it can be considered that there is no ghost target in the object detection region.

[0100] Step 502: For each first candidate region, in the object detection regions adjacent to the first candidate region and in the first candidate region, determine the region containing the number of target point cloud data that meets the preset quantity condition, and use it as the target region where the target object is located.

[0101] In implementation, after determining the first candidate region, the terminal can determine the object detection region adjacent to the first candidate region based on its location information. Then, the terminal can filter regions that meet a preset quantity condition from the first candidate region and its adjacent object detection regions. The preset quantity condition can be a maximum quantity. Specifically, the terminal can compare the number of target point cloud data within the first candidate region with the number of target point cloud data in each of the adjacent object detection regions, and select the region containing the most target point cloud data as the target region where the target object is located. For example, if the object detection region is a grid region, and the first candidate region (such as...) is determined... Figure 6 If the grid to be tested is shown, the terminal can determine the eight adjacent grid regions (e.g., the grid to be tested shown) based on the location information of each grid region. Figure 6 The test grid is divided into eight neighboring grids, each corresponding to an object detection region. The terminal can then select the region containing the most target point cloud data within these nine grid regions (including the test grid and the eight neighboring grids) as the target region for the target object. In this example, the neighboring grid region to the right of the first candidate region (the test grid) contains the most target point cloud data (15 points), so this neighboring grid region can be selected as the target region for the target object.

[0102] In this embodiment, after selecting the first candidate region based on confidence level, the region containing the most target point cloud data among the first candidate region and its adjacent regions is then selected as the region where the target object is located. On one hand, if the target object is large, or if it crosses from one object detection region to an adjacent object detection region within a multi-frame detection period, it may result in the simultaneous detection of the target object's point cloud data in multiple object detection regions. On the other hand, the location information of the multipath reflection point cloud data of the target object may exhibit location expansion, such as the actual lateral distance of the target being between -5m and -2m, while the point cloud data detected by the radar is between -6m and -1m. Therefore, further filtering of the target region based on the number of target point cloud data in the first candidate region and adjacent regions addresses the target distance expansion problem and prevents the same target from being detected simultaneously in adjacent regions. This allows for more accurate detection of the target object's location, improving target detection accuracy and reducing the false detection rate.

[0103] In one embodiment, such as Figure 7 As shown, the process of determining the target area where the target object is located in step 204 specifically includes the following steps:

[0104] Step 701: The object detection region with an object detection confidence level greater than or equal to a preset confidence threshold is determined as the second candidate region where the target object is located.

[0105] In practice, the terminal can determine the object detection region with an object detection confidence level greater than or equal to a preset confidence threshold as the second candidate region where the target object is located.

[0106] In some examples, the terminal may execute steps 501 and 502 and determine the region containing the most target point cloud data among the first candidate region and the neighboring object detection regions of the first candidate region as the second candidate region.

[0107] Step 702: In each second candidate region, determine the second candidate region pair that satisfies the similarity condition between the multipath target and the real target based on the feature information of the target point cloud data in each second candidate region.

[0108] In implementation, the terminal can determine pairs of second candidate regions that satisfy the similarity condition between the multipath target and the real target based on the feature information of the target point cloud data within each second candidate region. That is, the feature information of the target point cloud data contained in the two second candidate regions that make up the second candidate region pair will satisfy the similarity condition between the multipath target and the real target. The similarity condition between the multipath target and the real target can be set based on the feature information of the point cloud data of the real target and the multipath point cloud data of the multipath target corresponding to that real target. Therefore, in a pair of second candidate regions that satisfy the similarity condition between the multipath target and the real target, one second candidate region will contain the real target, and the other second candidate region will contain the multipath target corresponding to that real target. For example, the similarity condition between the multipath target and the real target can be that the velocity difference and longitudinal distance of the target point cloud data in the two second candidate regions satisfy threshold conditions, and the azimuth angles are symmetrical, that is, the azimuth angles are in opposite directions (e.g., one positive and one negative), and the difference in the azimuth angles satisfies the threshold condition. In one example, the terminal can calculate the average velocity, maximum velocity, and minimum velocity of the point cloud corresponding to each second candidate region, as well as the azimuth and position information of the point cloud data with the strongest point cloud intensity (referred to as the strongest point cloud), based on the feature information of the target point cloud data within each second candidate region, such as point cloud velocity, point cloud intensity, and azimuth. Then, the terminal can perform pairwise comparisons of the average velocity, maximum velocity, minimum velocity, azimuth, and position information of the strongest point cloud in each second candidate region. If two second candidate regions exist whose differences in average velocity, maximum velocity, minimum velocity, and longitudinal distance between the strongest point cloud and the target point cloud are all less than the corresponding preset thresholds, and whose azimuths are symmetrical, then these two second candidate regions are determined to be a pair of second candidate regions that satisfy the similarity condition between the multipath target and the real target.

[0109] Step 703: For the second candidate region pair, the second candidate region containing the most target point cloud data is selected as the target region where the target object is located.

[0110] In implementation, if the terminal identifies a pair of second candidate regions that satisfy the similarity condition between the multipath target and the real target in each second candidate region, the number of target point cloud data in the two second candidate regions of the pair can be compared, and the second candidate region containing the most target point cloud data can be identified as the target region where the target object is located. It can be assumed that the other second candidate region with fewer target point cloud data contains the multipath target corresponding to the target object (real target), meaning that there is no real target object in that region.

[0111] In some embodiments, if there are no other second candidate regions in each second candidate region that satisfy the similarity condition between the target second candidate region and the real target, the terminal can use the target second candidate region as the target region where the target object is located. Specifically, the terminal can perform similarity judgment on each pair of second candidate regions. The specific judgment method is detailed in the description of step 702 above, and will not be repeated here. If the terminal determines that there are no other second candidate regions that satisfy the similarity condition between the target and the real target for a certain second candidate region, the terminal can use the second candidate region as the target region where the target object is located. This situation may be that the point cloud data of the multipath target and the real target are detected in the same area, or only the multipath target is detected and the point cloud data of the occluded real target is not detected. In this case, the occluded target can be detected by accumulating multipath point clouds, so as to detect the ghosting target relatively early and improve driving safety.

[0112] In one possible implementation, the target detection area can be divided into grids to obtain multiple object detection areas (grid areas), and the determined second candidate areas are then the grid areas. The terminal can then traverse the second candidate areas within a certain vertical distance step, comparing whether there are pairs of second candidate areas within each vertical distance step that satisfy the similarity condition between the multipath target and the real target. For example, if the grid area size is 2m x 2m and the vertical distance step is set to 4m, then the second candidate areas can be traversed sequentially from near to far, with vertical distances of 0–4m, 4–8m, 8–12m, ..., 56–60m, that is, the similarity of the second candidate areas in every two rows of grid areas can be judged. The specific vertical distance step can be set according to the situation. A smaller step will lead to increased computation and even difficulty in accurately filtering out various multipath targets, resulting in a high false detection rate. A larger step will lead to some targets being incorrectly suppressed when there are many ghost targets in the environment, resulting in missed detections. In some examples, the vertical distance step can be set to 4 meters.

[0113] Specifically, if only one second candidate region is contained within a vertical distance step, for example, if the vertical distance of only one second candidate region is within 0-4m, then that second candidate region can be identified as the target region where the target object is located. If two or more second candidate regions are contained within a vertical distance step, then pairwise similarity judgment can be performed on the second candidate regions. If a pair of second candidate regions that satisfy the similarity condition between the multipath target and the real target is determined, then the second candidate region containing the most target point cloud data among the two second candidate regions is identified as the target region where the target object is located; if a second candidate region (the target second candidate region) does not have any other second candidate regions that satisfy the similarity condition between the multipath target and the real target, then the target second candidate region is identified as the target region where the target object is located. For example, if two rows of grid regions contain a pair of second candidate regions that satisfy the similarity condition between the multipath target and the real target (such as...), then... Figure 8 Using grids 1 and 2 as shown, the region containing the most target point cloud data in these two second candidate regions can be considered the region where the target object is located. In this example, grid 1 contains more target point cloud data (15 points) than grid 2 contains (10 points), so grid 1 is considered the target region where the target object is located, and it is assumed that there is no real target object in grid 2.

[0114] In this embodiment, based on the feature information of the target point cloud data within each second candidate region, a pair of second candidate regions that satisfy the similarity condition between the multipath target and the real target is determined. The second candidate region containing the most target point cloud data within the second candidate region pair is then selected as the target region. That is, when both a real target and its corresponding multipath target are detected simultaneously, and the two are located in different regions, the region containing the multipath target is removed, and only the region containing the real target is retained. Therefore, while ensuring target detection sensitivity and timeliness, detection accuracy can be further improved.

[0115] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0116] Based on the same inventive concept, this application also provides a target detection apparatus for implementing the target detection method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more target detection apparatus embodiments provided below can be found in the limitations of the target detection method described above, and will not be repeated here.

[0117] In one embodiment, such as Figure 9 As shown, a target detection device 900 is provided, including: an acquisition module 901, a first determination module 902, a screening module 903, and a second determination module 904, wherein:

[0118] The acquisition module 901 is used to acquire a point cloud dataset obtained by detecting the driving environment of the target vehicle; the point cloud dataset contains multiple frames of point cloud data.

[0119] The first determining module 902 is used to determine the point cloud data within the detection area of ​​each object based on the position information of each point cloud data in the point cloud dataset.

[0120] The filtering module 903 is used to filter out target point cloud data that meets the point cloud data feature conditions of the target object from the point cloud data in the point cloud data of each object detection area based on the feature information of the point cloud data in each object detection area.

[0121] The second determining module 904 is used to determine the object detection confidence level corresponding to each object detection area based on the number of target point cloud data in each object detection area, and to determine the target area where the target object is located based on the object detection area where the object detection confidence level is greater than or equal to a preset confidence threshold.

[0122] In one embodiment, the first determining module 902 is specifically used to: divide the target detection area of ​​the driving environment into a grid to obtain multiple object detection areas; match the position information of each point cloud data in the point cloud dataset with the position information of each object detection area to determine the point cloud data within each object detection area.

[0123] In one embodiment, the second determining module 904 is specifically configured to: for each object detection region, determine the number of target point cloud data within the object detection region as the object detection confidence of the object detection region, and obtain the confidence threshold corresponding to the object detection region; the confidence threshold corresponding to the object detection region is negatively correlated with the distance between the object detection region and the target vehicle; if the object detection confidence of the object detection region is greater than or equal to the confidence threshold corresponding to the object detection region, determine the object detection region as the target region where the target object is located.

[0124] In one embodiment, the second determining module 904 is specifically used to: determine the confidence threshold corresponding to the object detection area based on the distance between the object detection area and the target vehicle, in a pre-established correspondence between distance and confidence threshold.

[0125] In one embodiment, the second determining module 904 is specifically used to: determine the grid area corresponding to the object detection area based on the location information of the object detection area; and obtain the confidence threshold corresponding to the object detection area from the pre-established correspondence between the grid area and the confidence threshold.

[0126] In one embodiment, the second determining module 904 is specifically used to: determine the object detection region with an object detection confidence level greater than or equal to a preset confidence threshold as the first candidate region where the target object is located; for each first candidate region, determine the region containing the number of target point cloud data that meets the preset quantity condition among the object detection regions adjacent to the first candidate region and the first candidate region, and use it as the target region where the target object is located.

[0127] In one embodiment, the second determining module 904 is specifically used to: determine the object detection region with an object detection confidence level greater than or equal to a preset confidence threshold as the second candidate region where the target object is located; in each second candidate region, determine the second candidate region pair that satisfies the similarity condition between the multipath target and the real target based on the feature information of the target point cloud data in each second candidate region; for the second candidate region pair, take the second candidate region containing the most target point cloud data as the target region where the target object is located.

[0128] In one embodiment, the device further includes a third determining module, configured to, if in each second candidate region there are no other second candidate regions that satisfy the similarity condition between the multipath target and the real target with the target second candidate region, then regard the target second candidate region as the target region where the target object is located.

[0129] Each module in the aforementioned target detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0130] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a target detection method. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0131] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0132] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0133] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0134] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0135] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0136] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0137] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0138] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A target detection method characterized by, The method includes: A point cloud dataset is obtained by detecting the driving environment of the target vehicle; the point cloud dataset contains multiple frames of point cloud data. Based on the location information of each point cloud data in the point cloud dataset, determine the point cloud data within the detection area of ​​each object; Based on the feature information of the point cloud data in each object detection area, target point cloud data that meets the feature conditions of the target object is selected from the point cloud data in each object detection area. For each object detection region, the number of target point cloud data within the object detection region is determined as the object detection confidence of the object detection region, and a confidence threshold corresponding to the object detection region is obtained; the confidence threshold corresponding to the object detection region is negatively correlated with the distance between the object detection region and the target vehicle; If the object detection confidence level in the object detection region is greater than or equal to the confidence threshold corresponding to the object detection region, the object detection region is determined as the target region where the target object is located.

2. The method according to claim 1, characterized in that, The step of determining the point cloud data within the detection area of ​​each object based on the location information of each point cloud data in the point cloud dataset includes: The target detection area of ​​the driving environment is divided into grids to obtain multiple object detection areas; The location information of each point cloud data in the point cloud dataset is matched with the location information of each object detection region to determine the point cloud data within each object detection region.

3. The method according to claim 1, characterized in that, Based on the object detection regions where the object detection confidence level is greater than or equal to a preset confidence threshold, the target region where the target object is located is determined, including: The object detection region with an object detection confidence level greater than or equal to a preset confidence threshold is determined as the first candidate region where the target object is located; For each first candidate region, in the object detection regions adjacent to the first candidate region and the first candidate region, a region containing a number of target point cloud data that meets a preset quantity condition is determined as the target region where the target object is located.

4. The method according to claim 1, characterized in that, Based on the object detection regions where the object detection confidence level is greater than or equal to a preset confidence threshold, the target region where the target object is located is determined, including: The object detection region with an object detection confidence score greater than or equal to a preset confidence threshold is determined as the second candidate region where the target object is located; In each of the second candidate regions, a pair of second candidate regions that meet the similarity condition between the multipath target and the real target is determined based on the feature information of the target point cloud data in each of the second candidate regions. For the second candidate region pair, the second candidate region containing the most target point cloud data is taken as the target region where the target object is located.

5. The method according to claim 4, characterized in that, The method further includes: If there are no other second candidate regions in each of the second candidate regions that satisfy the similarity condition between the multipath target and the real target, then the second candidate region of the target is taken as the target region where the target object is located.

6. The method according to claim 1, characterized in that, The point cloud data feature conditions of the target object include one or more of the following: point cloud height condition, point cloud intensity condition, and point cloud velocity condition.

7. A target detection device, characterized in that, The device includes: The acquisition module is used to acquire a point cloud dataset obtained by detecting the driving environment of the target vehicle; the point cloud dataset contains multiple frames of point cloud data. The first determining module is used to determine the point cloud data within the detection area of ​​each object based on the location information of each point cloud data in the point cloud dataset. The filtering module is used to filter out target point cloud data that meets the point cloud data feature conditions of the target object from the point cloud data in each object detection area based on the feature information of the point cloud data in each object detection area; the confidence threshold corresponding to the object detection area is negatively correlated with the distance between the object detection area and the target vehicle; The second determining module is used to determine the number of target point cloud data in each object detection region as the object detection confidence of the object detection region, and to obtain the confidence threshold corresponding to the object detection region; the confidence threshold corresponding to the object detection region is negatively correlated with the distance between the object detection region and the target vehicle; If the object detection confidence level in the object detection region is greater than or equal to the confidence threshold corresponding to the object detection region, the object detection region is determined as the target region where the target object is located.

8. The apparatus according to claim 7, characterized in that, The first determining module is specifically used for: The target detection area of ​​the driving environment is divided into grids to obtain multiple object detection areas; The location information of each point cloud data in the point cloud dataset is matched with the location information of each object detection region to determine the point cloud data within each object detection region.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images