Trailer queue tracking method, device, computer equipment and storage medium

Abstract

The present invention provides a trailer queue tracking method, device, computer equipment and storage medium, and relates to the field of laser radar perception technology. The method includes: pre-processing the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue includes multiple targets, and each target corresponds to a subset of the point cloud data; according to the subset of point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames, determine the motion state of each target; according to the motion state of multiple targets in each candidate queue, determine whether each candidate queue is a trailer queue; if so, create a trailer queue model for the trailer queue according to the subset of point cloud data corresponding to multiple targets in the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames. The present invention can track the motion state of the trailer queue more accurately and stably.

Description

Technical Field

[0001] The present invention relates to the field of laser radar sensing technology, and in particular to a trailer queue tracking method, device, computer equipment and storage medium. Background Art

[0002] At present, using lidar as a sensor for autonomous vehicles to detect obstacles around the autonomous vehicles is a common obstacle detection solution.

[0003] In large-scale logistics scenarios, trailers are often connected to multiple trailers to form a trailer queue. Since the point cloud data generated by the lidar itself is sparse and there is occlusion between the trailers and the trailers during the movement of the trailer queue, the autonomous driving vehicle cannot accurately judge the movement state of the trailer queue, which is not conducive to tracking the trailer queue.

[0004] In this case, the autonomous driving vehicle may be unable to grasp the movement status of the trailer queue in real time, resulting in unnecessary emergency braking or failure to avoid in time, causing the autonomous driving vehicle to collide with the trailer queue. Summary of the invention

[0005] The purpose of the present invention is to provide a trailer queue tracking method, device, computer equipment and storage medium to solve the above-mentioned deficiencies in the prior art, so as to track the trailer queue in real time.

[0006] To achieve the above purpose, the technical solution adopted by the embodiment of the present invention is as follows:

[0007] In a first aspect, an embodiment of the present invention provides a method for tracking a trailer queue, comprising:

[0008] Preprocessing point cloud data collected by a laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue including a plurality of targets, each target corresponding to a subset of the point cloud data;

[0009] Determine the motion state of each target according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first multiple consecutive frames;

[0010] Determining whether each candidate queue is a trailer queue according to the motion states of the multiple targets in each candidate queue;

[0011] If so, a trailer queue model is created for the trailer queue based on a subset of the point cloud data corresponding to the multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second consecutive multiple frames, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets of the trailer queue.

[0012] Optionally, the preprocessing of the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue includes:

[0013] Clustering the point cloud data collected by the laser radar to obtain multiple cluster targets;

[0014] From the multiple cluster targets, determine a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain the candidate queue according to the plurality of candidate targets.

[0015] Optionally, before determining, from the multiple clustering targets, a distance between adjacent clustering targets is less than a preset distance, and a plurality of continuous candidate targets whose number is greater than or equal to a preset number threshold, and obtaining the candidate queue according to the multiple candidate targets, the method further includes:

[0016] Filter out cluster targets whose shape and size do not meet the preset size condition from the multiple cluster targets to obtain multiple initial targets;

[0017] The step of determining, from the plurality of cluster targets, a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtaining the candidate queue according to the plurality of candidate targets comprises:

[0018] From the multiple initial targets, determine the multiple consecutive candidate targets whose distance between adjacent initial targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain the candidate queue based on the multiple candidate targets.

[0019] Optionally, determining whether each candidate queue is a trailer queue according to the motion states of the multiple targets in each candidate queue includes:

[0020] Determining whether the motion states of the multiple targets in each candidate queue in the first continuous multiple frames are consistent according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames;

[0021] Inputting the point cloud data subset corresponding to the multiple targets of each candidate queue in the last frame of point cloud data in the first continuous multiple frames into a pre-trained trailer recognition model to determine whether the multiple targets of each candidate queue are recognized as trailers;

[0022] If the motion states of the multiple targets of each candidate queue in the first continuous multiple frames are consistent, and the multiple targets of each candidate queue are identified as trailers, then determining that each candidate queue is a trailer queue;

[0023] The trailer recognition model is a model pre-trained using sample point cloud data labeled with trailer information.

[0024] Optionally, creating a trailer queue model for the trailer queue according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames includes:

[0025] Determining whether multiple targets of the trailer queue corresponding to a subset of the point cloud data are all observed in the second plurality of consecutive frames of collected point cloud data;

[0026] If so, the trailer queue model is created according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames.

[0027] Optionally, after creating the trailer queue model according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames, the method further includes:

[0028] The trailer queue model is updated according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in each frame after the second consecutive multiple frames.

[0029] Optionally, the method further includes:

[0030] The motion states of multiple targets in the trailer queue are corrected according to the motion state of the trailer queue model.

[0031] The second method, an embodiment of the present invention further provides a trailer queue tracking device, the device comprising:

[0032] A preprocessing module, used to preprocess the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue including multiple targets, each target corresponding to a subset of the point cloud data;

[0033] A motion state calculation module, which determines the motion state of each target according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames;

[0034] A trailer queue determination module, configured to determine whether each candidate queue is a trailer queue according to the motion states of the multiple targets in each candidate queue;

[0035] A model creation module is used to create a trailer queue model for the trailer queue according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames if each of the queues is a trailer queue, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets of the trailer queue.

[0036] Optionally, the preprocessing module includes:

[0037] A clustering unit, used for clustering the point cloud data collected by the laser radar to obtain a plurality of clustering targets;

[0038] The target screening unit is used to determine, from the multiple cluster targets, a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain the candidate queue based on the plurality of candidate targets.

[0039] Optionally, before the target screening unit, the device further includes:

[0040] A filtering unit, used to filter out cluster targets whose shape and size do not meet a preset size condition from the plurality of cluster targets, to obtain a plurality of initial targets;

[0041] The screening unit is specifically used to determine, from the multiple initial targets, a plurality of continuous candidate targets whose distance between adjacent initial targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain the candidate queue based on the multiple candidate targets.

[0042] Optionally, the trailer queue determination module includes:

[0043] a motion state judgment unit, configured to determine whether the motion states of the multiple targets in each candidate queue in the first continuous multiple frames are consistent according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames;

[0044] a trailer judgment unit, configured to input the point cloud data subset corresponding to the multiple targets of each candidate queue in the last frame of point cloud data in the first continuous multiple frames into a pre-trained trailer recognition model to determine whether the multiple targets of each candidate queue are recognized as trailers;

[0045] a trailer queue determining unit, configured to determine each candidate queue as a trailer queue if the motion states of the multiple targets of each candidate queue in the first continuous multiple frames are consistent and the multiple targets of each candidate queue are identified as trailers;

[0046] The trailer recognition model is a model pre-trained using sample point cloud data labeled with trailer information.

[0047] Optionally, the model creation module includes:

[0048] a data observation judgment unit, configured to judge whether the plurality of targets of the trailer queue corresponding to the subset of the point cloud data are all observed in the second plurality of consecutive frames of collected point cloud data;

[0049] A model creation unit is used to create the trailer queue model according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames.

[0050] Optionally, after the model creation unit, the device further includes:

[0051] A model updating unit is used to update the trailer queue model according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in each frame after the second continuous multiple frames.

[0052] Optionally, the device further comprises:

[0053] A correction model is used to correct the motion states of multiple targets in the trailer queue according to the motion states of the trailer queue model.

[0054] In a third aspect, an embodiment of the present invention further provides a computer device, comprising: a processor and a memory, wherein the memory stores program instructions executable by the processor, and when the computer device is running, the processor executes the program instructions stored in the memory to perform the steps of the trailer queue tracking method as described in any of the above embodiments.

[0055] In a fourth aspect, an embodiment of the present invention further provides a storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to execute the steps of the trailer queue tracking method as described in any of the above embodiments.

[0056] The beneficial effects of the present invention are:

[0057] The present invention provides a trailer queue tracking method, device, computer equipment and storage medium. The method obtains at least one candidate queue by preprocessing the point cloud data collected by the laser radar mounted on the autonomous driving vehicle, each candidate queue includes multiple targets, each target corresponds to a subset of the point cloud data; the motion state of each target is determined according to the subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames; according to the motion states of the multiple targets of each candidate queue, it is determined whether each candidate queue is a trailer queue; if so, a trailer queue model is created for the trailer queue according to the subset of the point cloud data corresponding to the multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets of the trailer queue. The scheme of the present invention establishes a candidate queue through the point cloud data collected by the laser radar, and determines the trailer queue according to the motion states of the multiple targets in the candidate queue, so as to create a trailer queue model for the trailer queue, so as to track the motion state of the trailer queue more accurately and stably according to the trailer queue model, and avoid collision between the autonomous driving vehicle and the trailer queue. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 A schematic flow chart of a first trailer queue tracking method provided by an embodiment of the present invention;

[0060] Figure 2 A schematic flow chart of a second trailer queue tracking method provided by an embodiment of the present invention;

[0061] Figure 3 A schematic flow chart of a third trailer queue tracking method provided by an embodiment of the present invention;

[0062] Figure 4 A schematic flow chart of a fourth trailer queue tracking method provided by an embodiment of the present invention;

[0063] Figure 5 An experimental effect diagram of a trailer queue tracking method provided by an embodiment of the present invention;

[0064] Figure 6 A schematic diagram of the structure of a trailer queue tracking device provided by an embodiment of the present invention;

[0065] Figure 7A schematic diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0067] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] In the description of the present invention, it should be noted that if the terms "upper", "lower", etc. appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the drawings, or is the orientation or position relationship in which the product of the application is usually placed when used. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0069] In addition, the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0070] It should be noted that, in the absence of conflict, the features in the embodiments of the present invention may be combined with each other.

[0071] Figure 1 A schematic diagram of a flow chart of a first trailer queue tracking method provided by an embodiment of the present invention; Figure 1 As shown, the method includes:

[0072] S10: Preprocessing the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue including multiple targets, and each target corresponding to a subset of the point cloud data.

[0073] Specifically, the autonomous driving vehicle moves in a preset unmanned driving scene. The autonomous driving vehicle is equipped with a lidar, which is used to emit laser within a preset range where the autonomous driving vehicle is located, and collect point cloud data reflected by the laser on the surfaces of other objects within the preset range. The point cloud data includes information such as the orientation and distance of the object surface relative to the lidar.

[0074] In one possible implementation, in addition to using a lidar to collect point cloud data, a camera device can also be installed on the autonomous driving vehicle to take pictures of other objects in a preset unmanned driving scene, determine pixel data from the captured images, and pre-process the pixel data into at least one candidate queue.

[0075] The collected point cloud data is preprocessed, and the point cloud data is divided into multiple subsets of point cloud data according to information such as the distance between the point cloud data, each subset of point cloud data constitutes a target, and multiple targets constitute a candidate queue. In some embodiments, it can be determined whether each target constitutes at least one candidate queue based on the relative position of each target (each point cloud data) and the distance between each target (each point cloud data) and the autonomous driving vehicle.

[0076] S20: Determine the motion state of each target according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames.

[0077] Specifically, the laser radar emits laser within a preset range where the autonomous driving vehicle is located in the first multiple consecutive frames to collect multiple frames of point cloud data, and the point cloud data collected for each frame is preprocessed using the aforementioned S10 method to obtain a candidate queue for each frame.

[0078] For each target in the candidate queue, the moving distance of each target between each adjacent frames in the first continuous multiple frames is determined according to the subset of point cloud data corresponding to each target in each frame, and the time interval between each adjacent frames in the first continuous multiple frames is determined, and the motion state of each target between two adjacent frames is calculated according to the moving distance and the time interval, wherein the motion state of each target includes the motion speed and motion direction of each target.

[0079] S30: Determine whether each candidate queue is a trailer queue according to the motion states of the multiple targets in each candidate queue.

[0080] Specifically, according to the above S20, the motion states of the multiple targets in each candidate queue are obtained, and it is determined whether the motion states of the multiple targets in each candidate queue are consistent within the first continuous multiple frames, that is, whether the motion speeds and motion directions of the multiple targets in each candidate queue are consistent, so as to determine whether each candidate queue is a trailer queue, wherein the trailer queue is a queue formed by a trailer and at least one trailer moving in a preset unmanned driving scene. If the motion states of the multiple targets in a candidate queue are consistent within the first continuous multiple frames, the candidate queue is determined to be a trailer queue.

[0081] S40: If so, a trailer queue model is created for the trailer queue based on a subset of point cloud data corresponding to multiple targets in the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets in the trailer queue.

[0082] Specifically, if the candidate queue is determined to be a trailer queue through the above S30, according to the motion states of multiple targets of the trailer queue, a subset of point cloud data corresponding to multiple targets of the trailer queue is determined from the point cloud data collected by the laser radar in the second continuous multiple frames, and a preset queue model construction method is used to create a trailer queue model according to the subset of point cloud data corresponding to multiple targets of the trailer queue. For example, the queue model construction method can be: a subset of point cloud data corresponding to each target is represented by a rectangular frame, and multiple adjacent rectangular frames can constitute a trailer queue model for displaying the shape and motion state of the trailer queue, and the motion state of the trailer queue model is consistent with the motion state of multiple targets of the trailer queue.

[0083] The embodiment of the present invention provides a method for tracking a trailer queue, which pre-processes the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue includes multiple targets, and each target corresponds to a subset of the point cloud data; the motion state of each target is determined according to the subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames; according to the motion states of the multiple targets of each candidate queue, it is determined whether each candidate queue is a trailer queue; if so, a trailer queue model is created according to the subset of the point cloud data corresponding to the multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets of the trailer queue. The scheme of the embodiment of the present invention can establish multiple targets according to the point cloud data, determine whether the candidate queue is a trailer queue according to the motion state of the target, and create a trailer queue model according to the point cloud corresponding to the target of the trailer queue, so as to track the motion state of the trailer queue more accurately and stably through the trailer queue model, and avoid collision between the autonomous driving vehicle and the trailer queue.

[0084] Based on the above embodiment, the embodiment of the present invention further provides a method for tracking a trailer queue. Figure 2 A flow chart of a second trailer queue tracking method provided by an embodiment of the present invention is shown as follows: Figure 2 As shown, the above S10 includes:

[0085] S11: Clustering the point cloud data collected by the lidar to obtain multiple cluster targets.

[0086] Specifically, the laser emitted by the lidar irradiates the surface of the object and collects reflected point cloud data. The point cloud data is a plurality of discrete data points. The point cloud data with a distance less than a set threshold is clustered into the same cluster target to obtain multiple cluster targets.

[0087] S12: From the plurality of cluster targets, determine a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain a candidate queue according to the plurality of candidate targets.

[0088] Specifically, the trailer queue in the preset unmanned driving scene is composed of a trailer and at least one trailer. There is a fixed spacing distance between the trailer and the first trailer and between adjacent trailers to ensure that the trailer can move together with the trailer in the preset unmanned driving scene.

[0089] By calculating the spacing between adjacent clustering targets in a plurality of clustering targets, the spacing between any adjacent clustering targets is determined. According to the spacing between adjacent clustering targets, adjacent clustering targets with spacing less than a preset spacing are selected from the plurality of adjacent clustering targets, and the plurality of adjacent clustering targets with spacing less than the preset spacing are taken as a group of adjacent clustering targets, so as to select a plurality of groups of adjacent clustering targets with spacing less than the preset spacing from the plurality of clustering targets.

[0090] A trailer queue includes a trailer and at least one trailer. In order to filter out adjacent cluster targets corresponding to the trailer queue from multiple groups of adjacent cluster targets, it is first necessary to filter out at least one group of candidate targets from the multiple groups of adjacent cluster targets according to the number of the trailer and at least one trailer in the trailer queue. The number of cluster targets in each group of candidate targets is greater than or equal to a preset number threshold, and multiple candidate targets in each group of candidate targets constitute a candidate queue.

[0091] In an optional implementation, before the above S12, the method further includes:

[0092] Among the multiple cluster targets, the cluster targets whose shape and size do not meet the preset size conditions are filtered out to obtain multiple initial targets.

[0093] Specifically, in the preset unmanned driving scenario, in addition to the trailer queue, there may be other obstacles. The laser emitted by the lidar will also collect corresponding point cloud data when passing through the surface of other obstacles. In order to eliminate the interference of the point cloud data of other obstacles in creating the trailer queue model, the point cloud data of other obstacles needs to be filtered out.

[0094] In this embodiment, after clustering all the collected point cloud data to obtain multiple cluster targets, the cluster targets whose shape and size do not meet the preset size condition can be filtered out from the multiple cluster targets to obtain multiple initial targets. For example, the preset size condition can be set according to the size of the trailer.

[0095] Based on this, the above S12 includes:

[0096] From the multiple initial targets, determine a plurality of continuous candidate targets whose distance between adjacent initial targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain a candidate queue based on the plurality of candidate targets.

[0097] The trailer queue tracking method provided by the embodiment of the present invention obtains multiple cluster targets by clustering the point cloud data collected by the laser radar, and determines from the multiple cluster targets that the distance between adjacent cluster targets is less than the preset distance, and the number of multiple continuous candidate targets is greater than or equal to the preset number threshold, and obtains a candidate queue based on the multiple candidate targets. Through the embodiment of the present invention, adjacent cluster targets with too large distances and multiple continuous candidate targets with very few numbers can be regarded as targets that do not meet the requirements of the trailer queue. By filtering the cluster targets that do not meet the conditions, the candidate queue obtained based on the multiple candidate targets is more similar to the actual trailer queue, thereby improving the accuracy of creating the trailer queue model.

[0098] Based on the above embodiment, the embodiment of the present invention further provides a method for tracking a trailer queue. Figure 3 A schematic diagram of a flow chart of a third trailer queue tracking method provided by an embodiment of the present invention is shown as follows: Figure 3 As shown, the above S30 includes:

[0099] S31: Determine whether the motion states of multiple targets in each candidate queue in the first multiple consecutive frames are consistent based on a subset of point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first multiple consecutive frames.

[0100] Specifically, within the first multiple consecutive frames, a subset of point cloud data corresponding to each target in each frame is determined, and based on the subset of point cloud data corresponding to each target between two adjacent frames, the moving distance of each target between two adjacent frames is calculated. In combination with the time interval between two adjacent frames, the motion state between two adjacent frames is calculated, and it is determined whether the motion states of multiple targets between each adjacent multiple frames are consistent, thereby determining whether the motion states of multiple targets in each candidate queue are consistent within the first multiple consecutive frames.

[0101] S32: Inputting a subset of point cloud data corresponding to multiple targets of each candidate queue in the last frame of point cloud data in the first continuous multiple frames into a pre-trained trailer recognition model to determine whether the multiple targets of each candidate queue are recognized as trailers.

[0102] Specifically, the trailer identification model is pre-trained using a deep learning method, and can be a deep learning model that identifies the trailer based on point cloud data. During the training process, sample point cloud data pre-labeled with trailer information can be input into the initial trailer identification model, and the training parameters of the model are adjusted according to the output results of the initial trailer identification model until the output results of the model meet the requirements, thereby obtaining a trained trailer identification model.

[0103] From the point cloud data of the last frame of the first continuous multiple frames, a subset of the point cloud data corresponding to the multiple targets of each candidate queue is determined, and the subset of the point cloud data corresponding to the multiple targets is input into the trailer recognition model obtained by the above training. The trailer recognition model can be used to identify whether the multiple targets are trailers.

[0104] S33: If the motion states of the multiple targets of each candidate queue are consistent in the first multiple consecutive frames, and the multiple targets of each candidate queue are identified as trailers, each candidate queue is determined to be a trailer queue.

[0105] Specifically, if it is determined through the above S31 that the motion states of multiple targets of each candidate queue are consistent in the first multiple consecutive frames, and the trailer recognition model through the above S32 recognizes the multiple targets of the candidate queue as trailers, then the candidate queue is determined to be a trailer queue.

[0106] It should be noted that, since the trailer queue includes a trailer and at least one trailer, multiple targets are identified through the trailer identification model, the number of identified trailers is determined, and it is judged whether the number of identified trailers is the same as the number of trailers of the trailer queue in the preset unmanned driving scenario. If the number is the same and the movement state is consistent, the candidate queue is determined to be a trailer queue.

[0107] The trailer queue identification method provided by the embodiment of the present invention determines whether the motion states of multiple targets in each candidate queue are consistent in the first continuous multiple frames based on the subset of point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames, and inputs the subset of point cloud data corresponding to multiple targets in each candidate queue in the last frame of point cloud data in the first continuous multiple frames into a pre-trained trailer recognition model to determine whether the multiple targets in each candidate queue are identified as trailers. If the motion states of multiple targets in each candidate queue are consistent in the first continuous multiple frames, and the multiple targets in each candidate queue are identified as trailers, then each candidate queue is determined to be a trailer queue. Through the embodiment of the present invention, it is possible to determine whether a candidate queue is a trailer queue by judging the motion states of multiple targets in the candidate queue in the first continuous multiple frames and combining the recognition results of the trailer recognition model, so that the trailer queue can be accurately identified from the point cloud data.

[0108] Based on the above embodiment, the embodiment of the present invention further provides a method for tracking a trailer queue. Figure 4 A flowchart of a fourth trailer queue tracking method provided by an embodiment of the present invention is shown as follows: Figure 4 As shown, the above S40 includes:

[0109] S41: Determine whether a subset of point cloud data corresponding to multiple targets of the trailer queue is observed in the point cloud data collected in the second continuous multiple frames.

[0110] Specifically, after the trailer queue is determined from the point cloud data collected in the first continuous multiple frames, the subset of point cloud data corresponding to the multiple targets of the trailer queue in the first continuous multiple frames is observed to determine whether the point cloud data collected in the second continuous multiple frames includes the subset of point cloud data corresponding to the multiple targets.

[0111] S42: If yes, then create a trailer queue model according to a subset of point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames.

[0112] Specifically, if the point cloud data of each frame in the second continuous multiple frames includes a subset of point cloud data corresponding to multiple targets of the trailer queue, it is determined that the trailer queue is operating stably in the preset unmanned driving scene. From the point cloud data of the last frame of the second continuous multiple frames, a subset of point cloud data corresponding to the trailer queue is determined, and a preset queue model construction method is used to create a trailer queue model according to the subset of point cloud data corresponding to multiple targets of the trailer queue. The specific model creation process refers to the above S40 and is not described in detail here.

[0113] The trailer queue tracking method provided by the embodiment of the present invention determines whether a subset of point cloud data corresponding to multiple targets of the trailer queue is observed in the point cloud data collected in the second continuous multiple frames. If so, a trailer queue model is created based on the subset of point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames. Through the embodiment of the present invention, after the trailer queue is determined in the first continuous multiple frames, an accurate trailer queue model can be created when a subset of point cloud data corresponding to multiple targets of the trailer queue is observed in the second continuous multiple frames.

[0114] Based on the above embodiment, the embodiment of the present invention further provides a method for tracking a trailer queue. After S42, the method further includes:

[0115] The trailer queue model is updated according to a subset of point cloud data corresponding to multiple targets in the trailer queue in the point cloud data collected by the laser radar in each frame after the second consecutive multiple frames.

[0116] Specifically, the laser radar continuously emits laser after the second continuous multiple frames to collect multiple frames of point cloud data after the second continuous multiple frames, and determines a subset of point cloud data corresponding to multiple targets of the trailer queue from the point cloud data of each frame after the second continuous multiple frames, and uses a preset queue model construction method to create a trailer queue model according to the subset of point cloud data corresponding to multiple targets of the trailer queue in each frame, so as to continuously update the trailer queue model according to the motion state of the trailer queue in each frame after the second continuous multiple frames.

[0117] The trailer queue tracking method provided by the embodiment of the present invention updates the trailer queue model according to a subset of point cloud data corresponding to multiple targets in the trailer queue in each frame of point cloud data collected by the laser radar after the second continuous multiple frames. Through the embodiment of the present invention, the trailer queue model is continuously updated to continuously track the motion state of the trailer queue, thereby avoiding collision between the autonomous driving vehicle and the trailer queue.

[0118] Based on the above embodiment, an embodiment of the present invention further provides a method for tracking a trailer queue, the method further comprising:

[0119] The motion states of multiple targets in the trailer queue are corrected according to the motion states of the trailer queue model.

[0120] Specifically, the movement speed of the trailer queue model is calculated based on the movement distance and movement time of the trailer queue model in multiple consecutive frames, and the tangent direction of the curve of the trailer queue model is the movement direction of the trailer queue model. The same calculation method is used to calculate the movement state of multiple targets in the trailer queue. Due to the difference in the accuracy of the point cloud data collected by the laser radar through the emission of laser, the movement state of the multiple targets calculated is different from the overall movement state of the trailer queue model. Therefore, the movement state of the trailer queue model can be used to correct the movement state of the multiple targets in the trailer queue.

[0121] The trailer queue tracking method provided by the embodiment of the present invention corrects the motion states of multiple targets in the trailer queue according to the motion state of the trailer queue model. Through the embodiment of the present invention, the motion states of multiple targets in the trailer queue can be corrected to obtain a more accurate motion state of the target.

[0122] Based on the above embodiments, Figure 5 The experimental effect diagram of a trailer queue tracking method provided by an embodiment of the present invention is as follows: Figure 5 As shown in the figure, the ring wave is the laser emitted outward by the laser radar mounted on the autonomous driving vehicle. The left side is the trailer queue model obtained by the existing trailer queue tracking method, and the right side is the trailer queue model obtained by the trailer queue tracking method according to the embodiment of the present invention. It can be seen from the figure that the trailer queue model obtained by the trailer queue tracking method according to the embodiment of the present invention can simulate the motion state of the trailer queue more stably and accurately.

[0123] Based on the above embodiment, the embodiment of the present invention further provides a trailer queue tracking device. Figure 6 A schematic diagram of a trailer queue tracking device provided by an embodiment of the present invention is shown in FIG. Figure 6 As shown, the device comprises:

[0124] A preprocessing module 100 is used to preprocess the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue includes multiple targets, and each target corresponds to a subset of the point cloud data;

[0125] The motion state calculation module 200 determines the motion state of each target according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames;

[0126] The trailer queue determination module 300 is used to determine whether each candidate queue is a trailer queue according to the motion status of multiple targets in each candidate queue;

[0127] The model creation module 400 is used to create a trailer queue model for the trailer queue based on a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets of the trailer queue.

[0128] Optionally, the preprocessing module 100 includes:

[0129] A clustering unit is used to cluster the point cloud data collected by the laser radar to obtain multiple cluster targets;

[0130] The target screening unit is used to determine, from multiple cluster targets, a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain a candidate queue based on the plurality of candidate targets.

[0131] Optionally, before the target screening unit, the device further includes:

[0132] A filtering unit, used to filter out cluster targets whose shape and size do not meet a preset size condition from among the multiple cluster targets, to obtain multiple initial targets;

[0133] The screening unit is specifically used to determine, from multiple initial targets, a plurality of continuous candidate targets whose distance between adjacent initial targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain a candidate queue based on the multiple candidate targets.

[0134] Optionally, the trailer queue determination module 300 includes:

[0135] A motion state judgment unit, used to determine whether the motion states of multiple targets in each candidate queue in the first continuous multiple frames are consistent according to a subset of point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames;

[0136] A trailer queue determination unit, used for inputting a plurality of target corresponding point cloud data subsets of each candidate queue in the last frame of point cloud data in the first continuous multiple frames into a pre-trained queue recognition model to determine whether each candidate queue is recognized as a trailer queue;

[0137] a trailer queue determination unit, configured to determine each candidate queue as a trailer queue if the motion states of multiple targets of each candidate queue are consistent in the first plurality of consecutive frames and each candidate queue is identified as a trailer queue;

[0138] The queue recognition model is a model pre-trained using sample point cloud data labeled with trailer queues.

[0139] Optionally, the model creation module 400 includes:

[0140] A data observation judgment unit, used to judge whether a subset of point cloud data corresponding to multiple targets of the trailer queue is observed in the point cloud data collected in the second continuous multiple frames;

[0141] The model creation unit is used to create a trailer queue model according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames.

[0142] Optionally, after the model creation unit, the device further includes:

[0143] The model updating unit is used to update the trailer queue model according to a subset of point cloud data corresponding to multiple targets in the trailer queue in the point cloud data collected by the laser radar in each frame after the second continuous multiple frames.

[0144] Optionally, the device further comprises:

[0145] The correction model is used to correct the motion states of multiple targets in the trailer queue according to the motion states of the trailer queue model.

[0146] The above-mentioned device is used to execute the method provided by the aforementioned embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0147] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital singnal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0148] Figure 7 This is a schematic diagram of a computer device provided in an embodiment of the present invention. The computer device 500 includes: a processor 501 and a memory 502.

[0149] The memory 502 stores program instructions executable by the processor 501. When the computer device 500 is running, the processor 501 executes the program instructions stored in the memory 502 to perform the above method embodiment. The specific implementation method and technical effect are similar and will not be repeated here.

[0150] Optionally, the present invention also provides a program product, such as a computer-readable storage medium, comprising a program, which is used to execute the above method embodiment when executed by a processor.

[0151] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0152] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0153] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0154] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English: Read-Only Memory, abbreviated: ROM), random access memory (English: Random Access Memory, abbreviated: RAM), disk or optical disk and other media that can store program codes.

[0155] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for tracking a trailer queue, characterized in that: include: Preprocessing point cloud data collected by a laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue including a plurality of targets, each target corresponding to a subset of the point cloud data; Determine the motion state of each target according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first multiple consecutive frames; If the motion states of multiple targets in a candidate queue are consistent within the first multiple consecutive frames, determining the candidate queue to be a trailer queue; If one of the candidate queues is a trailer queue, a trailer queue model is created for the trailer queue based on a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second consecutive multiple frames, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets of the trailer queue.

2. The method according to claim 1, characterized in that The step of preprocessing the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue includes: Clustering the point cloud data collected by the laser radar to obtain multiple cluster targets; From the multiple cluster targets, determine a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain the candidate queue according to the plurality of candidate targets.

3. The method according to claim 2, characterized in that Before determining, from the plurality of cluster targets, a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtaining the candidate queue according to the plurality of candidate targets, the method includes: Filter out cluster targets whose shape and size do not meet the preset size condition from the multiple cluster targets to obtain multiple initial targets; The step of determining, from the plurality of cluster targets, a plurality of continuous candidate targets whose distance between adjacent cluster targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtaining the candidate queue according to the plurality of candidate targets comprises: From the multiple initial targets, determine the multiple consecutive candidate targets whose distance between adjacent initial targets is less than a preset distance and whose number is greater than or equal to a preset number threshold, and obtain the candidate queue based on the multiple candidate targets.

4. The method according to claim 1, characterized in that: If the motion states of multiple targets in a candidate queue are consistent within the first multiple consecutive frames, determining that the candidate queue is a trailer queue includes: Determining whether the motion states of the multiple targets in the candidate queue in the first continuous multiple frames are consistent according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames; Inputting the point cloud data subset corresponding to the multiple targets of each candidate queue in the last frame of point cloud data in the first continuous multiple frames into a pre-trained trailer recognition model to determine whether the multiple targets of each candidate queue are recognized as trailers; If the motion states of the multiple targets of the one candidate queue in the first consecutive multiple frames are consistent, and the multiple targets of the one candidate queue are identified as trailers, determining that the one candidate queue is a trailer queue; The trailer recognition model is a model pre-trained using sample point cloud data labeled with trailer information.

5. The method according to claim 1, characterized in that The step of creating a trailer queue model for the trailer queue according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second continuous multiple frames comprises: Determining whether multiple targets of the trailer queue corresponding to a subset of the point cloud data are all observed in the second plurality of consecutive frames of collected point cloud data; If so, the trailer queue model is created according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames.

6. The method according to claim 5, characterized in that After creating the trailer queue model according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the last frame of point cloud data in the second continuous multiple frames, the method further includes: The trailer queue model is updated according to a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in each frame after the second consecutive multiple frames.

7. The method according to claim 1, characterized in that The method further comprises: The motion states of multiple targets in the trailer queue are corrected according to the motion state of the trailer queue model.

8. A trailer queue tracking device, characterized in that: The device comprises: A preprocessing module, used to preprocess the point cloud data collected by the laser radar mounted on the autonomous driving vehicle to obtain at least one candidate queue, each candidate queue including multiple targets, each target corresponding to a subset of the point cloud data; A motion state calculation module, which determines the motion state of each target according to a subset of the point cloud data corresponding to each target in the point cloud data collected by the laser radar in the first continuous multiple frames; A trailer queue determination module, configured to determine if the motion states of multiple targets in a candidate queue are consistent within a first plurality of consecutive frames, and determine that the candidate queue is a trailer queue; A model creation module is used to create a trailer queue model for the trailer queue based on a subset of the point cloud data corresponding to multiple targets of the trailer queue in the point cloud data collected by the laser radar in the second consecutive multiple frames if the one candidate queue is a trailer queue, and the motion state of the trailer queue model is consistent with the motion state of the multiple targets of the trailer queue.

9. A computer device, characterized in that: include: A processor and a memory, wherein the memory stores program instructions executable by the processor, and when the computer device is running, the processor executes the program instructions stored in the memory to perform the steps of the trailer queue tracking method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the trailer queue tracking method according to any one of claims 1 to 7 are executed.

Images