A cone race track boundary classification method, system, and storage medium
By using a cone detection and deep learning classification model based on laser point clouds, the spatial distribution features of cones are extracted and aggregated, solving the problems of unreliability and complexity in cone track boundary classification in autonomous racing cars, and achieving more reliable and efficient boundary recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SOUTH CHINA UNIV OF TECH
- Filing Date
- 2023-03-03
- Publication Date
- 2026-06-16
AI Technical Summary
In existing technologies, the unreliability of image data quality and the complexity and unreliability of fusing image processing results with point cloud perception results lead to unreliable and highly complex classification of cone track boundaries for autonomous racing cars.
Based on the cone detection results using laser point clouds, a classification model is trained using deep neural networks and machine learning methods. Features of spatial distribution relationships and intermediate layer features of the cones are extracted, and then aggregated and post-processed to obtain the cone boundary category.
It achieves more reliable, accurate, and faster acquisition of cone track boundary information, solving the problems of unreliability and high complexity in existing technologies, and improving the development efficiency of unmanned racing cars.
Smart Images
Figure CN116433958B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of autonomous driving technology, and in particular to a cone track boundary classification method system, device and storage medium. Background Technology
[0002] In autonomous racing, the track is defined by cones of different colors arranged in a specific order. Classifying the left and right boundaries of these cones is a crucial technical aspect for the safe and reliable operation of autonomous vehicles. Since the left and right boundaries of the track correspond to red and blue cones respectively, current methods typically involve processing image data captured by the onboard cameras of the autonomous vehicle to obtain cone color information. This information is then fused with the cone spatial coordinate information obtained from point cloud data to ultimately determine the cone track boundaries.
[0003] However, the quality of image data acquired by cameras is unreliable in many situations, such as under certain weather conditions, lighting conditions, or when the vehicle is bumpy. Furthermore, fusing image processing results with point cloud perception results not only requires strict spatial calibration and temporal synchronization between the LiDAR and the camera, but also issues such as occlusion in the image can lead to errors and loss of fused information, introducing significant complexity and unreliability to autonomous driving systems.
[0004] It is evident that in existing technologies, the method of fusing image data and point cloud data cannot reliably and effectively classify cone track boundaries. Summary of the Invention
[0005] In order to at least partially solve one of the technical problems existing in the prior art, the present invention aims to provide a cone track boundary classification method system, device and storage medium.
[0006] The technical solution adopted in this invention is:
[0007] A cone track boundary classification method includes the following steps:
[0008] Obtain the cone detection results based on laser point clouds; where the cone detection results are the spatial coordinates [x,y,z] of each cone.
[0009] The cone detection results are used as input to a pre-established classification model to extract the spatial distribution relationship features and intermediate layer features of the cones;
[0010] The classification model aggregates the interrelationship features and intermediate features of the spatial distribution of cones, and obtains the classification result based on the aggregation result; wherein, the classification result is used to indicate the boundary category of each cone, and the boundary category includes left boundary, right boundary, and others.
[0011] Furthermore, obtaining the cone detection results based on laser point clouds includes:
[0012] The laser point cloud data is used as input to a pre-established target detection model, and the target detection model outputs the cone detection result.
[0013] The cone-shaped target detection model is obtained by training or parameter optimization using deep neural networks, statistical methods, or machine learning methods based on historical point cloud data and real labeled data.
[0014] Furthermore, the historical point cloud data is acquired by vehicle-mounted LiDAR, and the real labeled data is the spatial coordinates [x,y,z] of the cone point cloud, as well as its size [w,h,l].
[0015] Furthermore, the classification model is obtained through supervised machine learning based on multiple sets of cone spatial coordinate data.
[0016] Furthermore, the steps for training the classification model include:
[0017] In the process of machine learning on multiple sets of cone spatial coordinate data, the model parameters of the classification model are optimized using a loss function, which includes a cross-entropy loss function or a focus loss function.
[0018] Furthermore, the classification model includes a feature extraction module, a spatial distribution feature extraction module, an aggregation module, and a post-processing module;
[0019] The feature extraction module is used to extract the position features of the cone as intermediate layer features;
[0020] The spatial distribution feature extraction module is used to extract the interrelationship features of the spatial distribution of cones;
[0021] The aggregation module is used to aggregate the interrelationship features between the intermediate layer features and the spatial distribution features of the cone;
[0022] The post-processing module is used to process the aggregated features and output the classification results.
[0023] Furthermore, the feature extraction module is a multilayer perceptron (MLP); the spatial distribution feature extraction module consists of a multilayer perceptron (MLP) and a max pooling operation (MaxPool); the aggregation module consists of a concatenation operation (Concat); and the post-processing module consists of a multilayer perceptron (MLP) and a normalized exponential function (SoftMax).
[0024] Another technical solution adopted in this invention is:
[0025] A cone track boundary classification system includes:
[0026] The acquisition module is used to acquire the cone detection results based on laser point clouds; wherein, the cone detection result is the spatial coordinates [x,y,z] of each cone;
[0027] The feature extraction module is used to take the cone detection results as input to the pre-established classification model and extract the spatial distribution relationship features of the cones and the intermediate layer features;
[0028] The classification module is used to aggregate the interrelationship features and intermediate features of the spatial distribution of cones through the classification model, and obtain the classification result based on the aggregation result; wherein, the classification result is used to indicate the boundary category of each cone, and the boundary category includes left boundary, right boundary, and others.
[0029] Another technical solution adopted in this invention is:
[0030] A cone track boundary classification device includes:
[0031] At least one processor;
[0032] At least one memory for storing at least one program;
[0033] When the at least one program is executed by the at least one processor, the at least one processor implements the method described above.
[0034] Another technical solution adopted in this invention is:
[0035] A computer-readable storage medium storing a processor-executable program, which, when executed by a processor, performs the method described above.
[0036] The beneficial effects of this invention are as follows: This invention aggregates the interrelationship features and intermediate features of the spatial distribution of cones using a classification model, and obtains the classification result based on the aggregation result. Compared with the existing method of detecting cones by fusing images and laser point clouds, this invention can obtain cone track boundary information more reliably, accurately, and quickly, thereby solving the problems of unreliability and high complexity in existing cone track boundary classification schemes. This invention can be effectively applied to cone track boundary classification for autonomous racing and can be extended to other similar scenarios, improving the development efficiency of autonomous driving. Attached Figure Description
[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of the present invention or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0038] Figure 1 This is a flowchart of the steps of a cone track boundary classification method in an embodiment of the present invention;
[0039] Figure 2 This is a schematic diagram of the classification model in an embodiment of the present invention;
[0040] Figure 3 This is an example diagram illustrating the classification of cone track boundaries in an embodiment of the present invention;
[0041] Figure 4 This is a structural block diagram of a cone track boundary classification system according to an embodiment of the present invention;
[0042] Figure 5 This is another structural block diagram of a cone track boundary classification system in an embodiment of the present invention;
[0043] Figure 6 This is a structural block diagram of a cone track boundary classification device according to an embodiment of the present invention. Detailed Implementation
[0044] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.
[0045] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.
[0046] In the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.
[0047] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.
[0048] In existing technologies, autonomous racing car technology typically classifies cone track boundaries by fusing image processing results with point cloud perception results. However, in many cases, due to factors such as the susceptibility of image data quality to environmental influences, target occlusion issues in images, and the need for strict spatiotemporal synchronization between cameras and LiDAR, it is impossible to reliably and effectively classify cone track boundaries using the method of fusing image processing results with point cloud perception results.
[0049] To address the aforementioned problems in the prior art, this application provides a method, system, apparatus, and storage medium for classifying cone track boundaries to solve these problems.
[0050] In the technical solution provided in this application, the cone track boundary classification method obtains cone detection results based on laser point clouds. The cone detection results are used as input to a pre-established classification model to extract the spatial distribution relationship features and intermediate layer features of the cones. The classification model then performs aggregation and post-processing on the spatial distribution relationship features and intermediate features of the cones to obtain the classification results. The classification results are used to indicate the boundary category of each cone, namely left boundary, right boundary, and others. Compared with the existing method of fusion of images and laser point clouds for track cone detection, this method can obtain cone track boundary information more reliably, accurately, and quickly, thereby solving the problems of unreliability and high complexity in existing cone track boundary classification schemes.
[0051] like Figure 1 As shown, this embodiment provides a cone track boundary classification method, including the following steps:
[0052] S1. Obtain the cone detection results based on laser point clouds; where the cone detection results are the spatial coordinates [x,y,z] of each cone.
[0053] In this embodiment, the point cloud data can be acquired through an onboard LiDAR on the vehicle.
[0054] Extracting barrel detection results from point cloud data can be achieved in various ways, including extracting barrel spatial coordinate data from point cloud data through object detection. Specifically, this involves feeding point cloud data as input to a pre-trained object detection model, which then outputs the barrel spatial coordinate data. The barrel object detection model is obtained by training and parameter optimization using deep neural networks, statistical methods, or machine learning methods based on historical point cloud data and real-world labeled data.
[0055] As an optional implementation, historical point cloud data is collected by vehicle-mounted LiDAR, and the actual labeled data is the spatial coordinates [x,y,z] of the cone, as well as its size [w,h,l].
[0056] In specific applications, the spatial coordinate data of the cone can also be extracted through other methods, which are not specifically limited here.
[0057] S2. The cone detection results are used as input to the pre-established classification model to extract the spatial distribution relationship features of the cones and the intermediate layer features.
[0058] In this embodiment, the classification model is obtained by supervised machine learning based on multiple sets of cone space coordinates.
[0059] In the process of machine learning on multiple sets of cone space coordinates, the classification model is optimized using a loss function, which includes a cross-entropy loss function or a focus loss function.
[0060] As an optional implementation method, such as Figure 2 As shown, the classification model structure includes a feature extraction module, a spatial distribution feature extraction module, an aggregation module, and a post-processing module. The feature extraction module is used to extract the positional features of the cones, i.e., the intermediate layer features. The spatial distribution feature extraction module is used to extract the spatial distribution relationship features of the cones. The aggregation module is used to aggregate the intermediate layer features and the spatial distribution relationship features of the cones. The post-processing module is used to process the aggregated features and output the classification result.
[0061] As an optional implementation, the feature extraction module consists of multiple sets of Multi-Layer Perception (MLP); the spatial distribution feature extraction module consists of MLP and MaxPool; the aggregation module consists of concatenation operation; and the post-processing module consists of MLP and SoftMax operation.
[0062] S3. Aggregate the interrelationship features and intermediate features of the spatial distribution of the cones using the classification model, and obtain the classification result based on the aggregation result. The classification result indicates the boundary category of each cone, including left boundary, right boundary, and others.
[0063] In this embodiment, "others" refers to cones that are misidentified by the cone target detection model and cones that do not belong to the current driving track.
[0064] See Figure 3 , Figure 3 This is a schematic diagram of boundary classification in this embodiment. Figure 3 (a) is a diagram showing the placement of the track cones. Figure 3 (b) shows the classification results of the track cone boundaries, where 0 represents the left boundary, 1 represents the right boundary, and 2 represents the others.
[0065] like Figure 4 As shown, this embodiment also provides a cone track boundary classification system, including:
[0066] The acquisition module 31 is used to acquire the cone detection results based on laser point clouds; wherein, the cone detection results are the spatial coordinates [x,y,z] of each cone;
[0067] The feature extraction module 32 is used to provide the cone detection results as input to the pre-established classification model to extract the spatial distribution relationship features of the cones and the intermediate layer features;
[0068] The classification module 33 is used to aggregate the interrelationship features and intermediate features of the spatial distribution of cones through the classification model, and obtain the classification result based on the aggregation result; wherein, the classification result is used to indicate the boundary category of each cone, and the boundary category includes left boundary, right boundary, and others.
[0069] As a further optional implementation, the acquisition module 31 acquires the cone detection result based on the laser point cloud, including: providing the laser point cloud data as input to a pre-established target detection model, and the target detection model outputs the cone detection result.
[0070] As a further optional implementation, the cone target detection model is obtained by training or parameter optimization using deep neural networks, statistical methods, or machine learning methods based on historical point cloud data and real labeled data.
[0071] As an optional implementation, the historical point cloud data is collected by vehicle-mounted LiDAR, and the actual labeled data is the spatial coordinates [x,y,z] of the cone point cloud, as well as its size [w,h,l].
[0072] The feature extraction module 32 takes the cone detection result information as input to the pre-established classification model and extracts the spatial distribution relationship features and intermediate layer features of the cones, including: the classification model is obtained by supervised machine learning based on multiple sets of cone spatial coordinate data.
[0073] As a further optional implementation, during the machine learning process using multiple sets of cone spatial coordinate data, the parameters of the classification model are optimized using a loss function, wherein the loss function includes a cross-entropy loss function or a focus loss function.
[0074] As an optional implementation, the classification model includes a feature extraction module, a spatial distribution feature extraction module, an aggregation module, and a post-processing module. The feature extraction module is used to extract the positional features of the cones, i.e., the intermediate layer features; the spatial distribution feature extraction module is used to extract the spatial distribution relationship features of the cones; the aggregation module is used to aggregate the intermediate layer features and the spatial distribution relationship features of the cones; and the post-processing module is used to process the aggregated features and output the classification result.
[0075] Further, as an optional implementation, such as Figure 5 As shown, the cone track boundary classification system may also include:
[0076] The detection module 34 is used to provide laser point cloud data as input to a pre-established target detection model, and the target detection model outputs the cone detection result.
[0077] Through such Figure 4 Or such as Figure 5 The system shown includes a cone detection module 31 that acquires cone detection results based on laser point clouds. A feature extraction module 32 provides the cone detection results as input to a pre-established classification model, extracting spatial distribution relationship features and intermediate layer features of the cones. A classification module 33 aggregates and post-processes the spatial distribution relationship features and intermediate layer features of the cones using the classification model to obtain a classification result. The classification result indicates the boundary category of each cone: left boundary, right boundary, or others. Therefore, the system provided in this embodiment can effectively classify cone track boundaries, solving the problems of unreliability and high complexity in existing cone track boundary classification schemes.
[0078] like Figure 6 As shown, this embodiment also provides a cone track boundary classification device 400, including a processor 401 and at least one memory 402. The at least one memory 402 stores at least one machine-executable instruction, and the processor 401 executes the at least one machine-executable instruction to perform:
[0079] Obtain the cone detection results based on laser point clouds; where the detection results are the spatial coordinates [x,y,z] of each cone.
[0080] The cone detection results are used as input to a pre-established classification model to extract the spatial distribution relationship features of the cones and the intermediate layer.
[0081] The classification results are obtained by aggregating and post-processing the spatial distribution interrelationship features and intermediate features of the cones using a classification model; the classification results are used to indicate the boundary category of each cone, namely left boundary, right boundary, and others.
[0082] The processor executes at least one machine-executable instruction to obtain cone detection results based on laser point clouds, including: providing point cloud data as input to a pre-trained target detection model, and the target detection model outputting cone spatial coordinate data.
[0083] As a further optional implementation, the cone target detection model is obtained by training or parameter optimization using deep neural networks, statistical methods, or machine learning methods based on historical point cloud data and real labeled data.
[0084] As an optional implementation, the historical point cloud data is collected by vehicle-mounted LiDAR, and the actual labeled data is the spatial coordinates [x,y,z] of the cone point cloud, as well as its size [w,h,l].
[0085] The processor executes at least one machine-executable instruction to take the cone detection result information as input to a pre-established classification model, extract the spatial distribution relationship features of the cones and the intermediate layer, including: the classification model is obtained by supervised machine learning based on multiple sets of cone spatial coordinate data.
[0086] As a further optional implementation, during the machine learning process using multiple sets of cone spatial coordinate data, the parameters of the classification model are optimized using a loss function, wherein the loss function includes a cross-entropy loss function or a focus loss function.
[0087] As an optional implementation, the classification model includes a feature extraction module, a spatial distribution feature extraction module, an aggregation module, and a post-processing module. The feature extraction module is used to extract the positional features of the cones, i.e., the intermediate layer features; the spatial distribution feature extraction module is used to extract the spatial distribution relationship features of the cones; the aggregation module is used to aggregate the intermediate layer features and the spatial distribution relationship features of the cones; and the post-processing module is used to process the aggregated features and output the classification result.
[0088] The processor executes at least one machine-executable instruction to perform aggregation and post-processing on the spatial distribution interrelationship features and intermediate layer features of the cones through the classification model to obtain a classification result, including: the classification result is used to indicate the boundary category of each cone, namely left boundary, right boundary, and others.
[0089] According to the apparatus provided in this application embodiment, the spatial coordinate data of cones obtained from laser point clouds are input to a pre-established classification model. Features of the spatial distribution relationships and intermediate layer features of the cones are extracted. The extracted features of spatial distribution relationships and intermediate layer features are then aggregated and post-processed by the classification model to obtain a classification result. The classification result indicates the boundary category of each cone, namely left boundary, right boundary, and others. Therefore, the apparatus provided in this application embodiment can effectively classify cone track boundaries, solving the problems of unreliability and high complexity in existing cone track boundary classification schemes.
[0090] This application also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform... Figure 1 The method shown.
[0091] This embodiment also provides a storage medium storing a method embodiment of the present invention that can execute the provided method embodiment. Figure 1 The instructions or program for a cone track boundary classification method are shown. When the instructions or program are run, any combination of implementation steps of the method embodiment can be executed, and the method has the corresponding functions and beneficial effects.
[0092] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.
[0093] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.
[0094] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0095] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0096] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0097] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0098] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0099] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
[0100] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.
Claims
1. A method for classifying the boundaries of a cone-shaped race track, characterized in that, Includes the following steps: Obtain the cone detection results based on laser point clouds; where the cone detection result is the spatial coordinates of each cone. ; The cone detection results are used as input to a pre-established classification model to extract the spatial distribution relationship features and intermediate layer features of the cones; The classification model aggregates the interrelationship features and intermediate features of the spatial distribution of cones, and obtains the classification result based on the aggregation result; wherein, the classification result is used to indicate the boundary category of each cone, and the boundary category includes left boundary, right boundary, and others; The acquisition of cone detection results based on laser point clouds includes: The laser point cloud data is used as input to a pre-established target detection model, and the target detection model outputs the cone detection result. The cone target detection model is obtained by training or parameter optimization using deep neural networks, statistical methods, or machine learning methods based on historical point cloud data and real labeled data. The classification model includes a feature extraction module, a spatial distribution feature extraction module, an aggregation module, and a post-processing module; The feature extraction module is used to extract the position features of the cone as intermediate layer features; The spatial distribution feature extraction module is used to extract the interrelationship features of the spatial distribution of cones; The aggregation module is used to aggregate the interrelationship features between the intermediate layer features and the spatial distribution features of the cone; The post-processing module is used to process the aggregated features and output the classification results.
2. The cone track boundary classification method according to claim 1, characterized in that, The historical point cloud data was acquired by vehicle-mounted LiDAR, and the real labeled data is the spatial coordinates of the cone point cloud. and its size .
3. The cone track boundary classification method according to claim 1, characterized in that, The classification model is obtained through supervised machine learning based on multiple sets of cone spatial coordinate data.
4. The cone track boundary classification method according to claim 3, characterized in that, The steps for training a classification model include: In the process of machine learning on multiple sets of cone spatial coordinate data, the model parameters of the classification model are optimized using a loss function, which includes a cross-entropy loss function or a focus loss function.
5. The cone track boundary classification method according to claim 1, characterized in that, The feature extraction module is a multilayer perceptron; the spatial distribution feature extraction module consists of a multilayer perceptron and a max pooling operation; the aggregation module consists of a concatenation operation; and the post-processing module consists of a multilayer perceptron and a normalized exponential function.
6. A cone track boundary classification system, characterized in that, include: The acquisition module is used to acquire the cone detection results based on laser point clouds; wherein, the cone detection result is the spatial coordinates of each cone. ; The feature extraction module is used to take the cone detection results as input to the pre-established classification model and extract the spatial distribution relationship features of the cones and the intermediate layer features; The classification module is used to aggregate the interrelationship features and intermediate features of the spatial distribution of cones using the classification model, and obtain the classification result based on the aggregation result; wherein, the classification result is used to indicate the boundary category of each cone, and the boundary category includes left boundary, right boundary, and others; The acquisition of cone detection results based on laser point clouds includes: The laser point cloud data is used as input to a pre-established target detection model, and the target detection model outputs the cone detection result. The cone target detection model is obtained by training or parameter optimization using deep neural networks, statistical methods, or machine learning methods based on historical point cloud data and real labeled data. The classification model includes a feature extraction module, a spatial distribution feature extraction module, an aggregation module, and a post-processing module; The feature extraction module is used to extract the position features of the cone as intermediate layer features; The spatial distribution feature extraction module is used to extract the interrelationship features of the spatial distribution of cones; The aggregation module is used to aggregate the interrelationship features between the intermediate layer features and the spatial distribution features of the cone; The post-processing module is used to process the aggregated features and output the classification results.
7. A cone track boundary classification device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method of any one of claims 1-5.
8. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to perform the method as described in any one of claims 1-5.