Method for constructing unstructured road bump feature model and related device

Abstract

The present application relates to the technical field of vehicle environment perception, and discloses a method for constructing an unstructured road bump feature model and related devices. The method includes: constructing an initial road model based on a Gaussian function; randomly generating a plurality of initial road surface point cloud data based on the initial road model; classifying and clustering the initial road surface point cloud data according to the bump feature information of the initial road surface point cloud data to obtain a plurality of point cloud clusters of interest; fitting each point cloud cluster of interest into a form represented by a plurality of two-dimensional Gaussian functions to obtain an unstructured road bump feature model. By reconstructing the road surface in the embodiments of the present application, the perception accuracy of the road surface can be improved.

Description

Technical Field

[0001] The present application relates to the field of vehicle environment perception technology, and in particular to a method for constructing an unstructured road bump feature model and related equipment. Background Art

[0002] Environmental perception technology is one of the core technologies of intelligent vehicle technology. It obtains and analyzes sensor data to provide environmental information for decision-making, control and other technologies.

[0003] However, when driving on unstructured roads such as mining areas and off-road scenarios, intelligent vehicles need to face problems such as changing road types and significant slope fluctuations. Current environmental perception technology lacks accurate model expression of the road surface when performing road surface recognition, resulting in reduced perception accuracy. Summary of the Invention

[0004] The purpose of this application is to provide a method for constructing an unstructured road bump feature model and related equipment, which can improve the perception accuracy of the road surface by remodeling the road surface.

[0005] The present invention provides a method for constructing an unstructured road concave-convex feature model, comprising:

[0006] Construct an initial road model based on Gaussian function;

[0007] Based on the initial road model, randomly generating a plurality of initial road surface point cloud data; the initial road surface point cloud data includes randomly generated concave-convex feature information;

[0008] Classifying and clustering the initial road surface point cloud data according to the concave-convex feature information of the initial road surface point cloud data to obtain a plurality of point cloud clusters of interest; the point cloud clusters of interest comprising a plurality of point cloud data of interest of the same category, the point cloud data of interest being initial road surface point cloud data having concave-convex feature information of interest, the concave-convex feature corresponding to the concave-convex feature information of interest generating an interference intensity on driving that reaches a preset interference intensity threshold;

[0009] Each of the interest point cloud clusters is fitted into a form represented by multiple two-dimensional Gaussian functions to obtain an unstructured road concave-convex feature model.

[0010] In some embodiments, the initial road model is expressed as:

[0011]

[0012] Among them, F i The (x) function represents the road point cloud at position i along the y-axis direction, as well as are the x, y, and z coordinates of the point cloud, respectively.i , B i and C i are all Gaussian function parameter values, ω i (x) is a uniformly distributed random function, i = 1, 2,..., n, where n is the total number of random road surface point cloud data.

[0013] In some embodiments, randomly generating a plurality of initial road surface point cloud data based on the initial road model includes:

[0014] Using the initial road model to randomly generate a plurality of random road surface point cloud data with corresponding amplitude, range, and distribution position of concave and convex feature information;

[0015] According to the concave and convex feature information of the random road surface point cloud data, performing class annotation on the random road surface point cloud data to obtain the randomly road surface point cloud data after class annotation;

[0016] Adopting a point cloud three-dimensional rigid body translation method to splice the randomly road surface point cloud data after class annotation into point cloud data of a specified size to obtain the initial road surface point cloud data.

[0017] In some embodiments, classifying and clustering the initial road surface point cloud data according to the concave and convex feature information of the initial road surface point cloud data to obtain a plurality of point cloud clusters of interest includes:

[0018] Inputting the initial road surface point cloud data into a concave and convex feature semantic segmentation model to perform concave and convex feature class prediction on the initial road surface point cloud data to obtain a point cloud data class prediction result; the concave and convex feature semantic segmentation model is trained by a neural network model based on the PointNet++ semantic segmentation network;

[0019] According to the point cloud data class prediction result, selecting point cloud data from the initial road surface point cloud data to obtain a plurality of the point cloud data of interest;

[0020] Clustering each of the point cloud data of interest to obtain a plurality of the point cloud clusters of interest.

[0021] In some embodiments, before inputting the initial road surface point cloud data into the concave and convex feature semantic segmentation model to perform concave and convex feature class prediction on the initial road surface point cloud data to obtain a point cloud data class prediction result, it further includes:

[0022] Inputting the initial road surface point cloud data into the neural network model to perform concave and convex feature class prediction on the initial road surface point cloud data to obtain a point cloud data class prediction sample;

[0023] Determine the model loss information of the neural network model according to the true category of the initial road surface point cloud data and the predicted samples of the point cloud data categories.

[0024] Train the neural network model according to the model loss information, and obtain the bump feature semantic segmentation model when the neural network model meets the training end condition.

[0025] In some embodiments, the fitting of each of the point cloud clusters of interest to be represented in the form of a plurality of two-dimensional Gaussian functions to obtain an unstructured road bump feature model includes:

[0026] Represent the point cloud clusters of interest using a plurality of two-dimensional Gaussian functions, and use the least squares method to fit the amplitude, the center coordinates of the amplitude point, the decay coefficient, and the distribution range of the point cloud clusters of interest to obtain the unstructured road bump feature model.

[0027] In some embodiments, the expression of the unstructured road bump feature model is:

[0028]

[0029] where Z j (x, y) represents the height of the road surface, and any (x, y) coordinate corresponds to a height value. A j is the amplitude of the point cloud clusters of interest, μ xj is the x-axis coordinate of the amplitude point of the point cloud clusters of interest, μ yj is the y-axis coordinate of the amplitude point of the point cloud clusters of interest, c j is the decay coefficient, is the distribution range of the point cloud clusters of interest in the positive x-axis direction, is the distribution range of the point cloud clusters of interest in the negative x-axis direction, is the distribution range of the point cloud clusters of interest in the positive y-axis direction, is the distribution range of the point cloud clusters of interest in the negative y-axis direction, j = 1, 2,..., m, and m is the total number of the point cloud clusters of interest.

[0030] The embodiments of the present application further provide an unstructured road bump feature model construction device, including:

[0031] A first module for constructing an initial road model based on a Gaussian function;

[0032] A second module for randomly generating a plurality of initial road surface point cloud data based on the initial road model; the initial road surface point cloud data includes randomly generated bump feature information;

[0033] The third module is used to classify and cluster the initial road surface point cloud data according to the concavo-convex feature information of the initial road surface point cloud data, so as to obtain multiple point cloud clusters of interest; each point cloud cluster of interest contains multiple point cloud data of the same category, the point cloud data of interest is the initial road surface point cloud data with concavo-convex feature information of interest, and the concavo-convex feature corresponding to the concavo-convex feature information of interest reaches a preset interference intensity threshold for driving.

[0034] The fourth module is used to fit each of the point cloud clusters of interest into a form represented by multiple two-dimensional Gaussian functions to obtain an unstructured road concavo-convex feature model.

[0035] An embodiment of the present application further provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above-mentioned method for constructing an unstructured road concavo-convex feature model is implemented.

[0036] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the above-mentioned method for constructing an unstructured road concavo-convex feature model is implemented.

[0037] The beneficial effects of the present application: Using the initial road model constructed based on Gaussian functions, multiple initial road surface point cloud data containing random concavo-convex feature information are randomly generated. According to the concavo-convex feature information of the initial road surface point cloud data, the initial road surface point cloud data is classified and clustered to obtain multiple point cloud clusters of interest composed of point cloud data of interest that have a significant impact on driving. Then, each point cloud cluster of interest is fitted into a form represented by multiple two-dimensional Gaussian functions to obtain an unstructured road concavo-convex feature model. Since multiple initial road surface point cloud data are randomly generated using the initial road model, and then the point cloud data of interest that have a significant impact on driving are selected from them, and the unstructured road concavo-convex feature model is reconstructed using the point cloud clusters of interest composed of point cloud data of the same category, the concavo-convex features of the unstructured road surface can be accurately expressed, and the perception accuracy of the road surface can be improved. Description of the Drawings

[0038] Figure 1 is a flowchart of the method for constructing an unstructured road concavo-convex feature model provided by an embodiment of the present application.

[0039] Figure 2 is a flowchart of the specific method of step S102 provided by an embodiment of the present application.

[0040] Figure 3 is a flowchart of the specific method of step S103 provided by an embodiment of the present application.

[0041] Figure 4It is a flowchart of the method before step S301 provided by an embodiment of the present application.

[0042] Figure 5 It is a schematic structural diagram of an unstructured road bump feature model construction device provided by an embodiment of the present application.

[0043] Figure 6 It is a schematic hardware structure diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0044] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

[0045] It should be noted that although functional module division is performed in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown can be executed in a different order from the module division in the device or the order in the flowchart. Terms such as "first" and "second" in the description, claims and drawings are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing embodiments of this application and are not intended to limit this application.

[0047] Refer to Figure 1 , Figure 1 It is a flowchart of an unstructured road bump feature model construction method provided by an embodiment of the present application. In some embodiments of the present application, Figure 1 the method in Figure 1 may specifically include but is not limited to steps S101 to S104, which will be introduced in detail below with reference to

[0048] Step S101, construct an initial road model based on the Gaussian function.

[0049] In a specific embodiment, the expression of the initial road model is:

[0050]

[0051] where the F i (x) function represents the road surface point cloud at position i along the y-axis direction, and are the x, y, and z coordinates of the point cloud respectively, and A i , B i and Ci are all Gaussian function parameter values, ω i (x) is a uniformly distributed random function, i = 1, 2,..., n, where n is the total number of random road surface point cloud data.

[0052] Step S102: Based on the initial road model, randomly generate multiple initial road surface point cloud data.

[0053] Among them, the initial road surface point cloud data contains randomly generated concave and convex feature information.

[0054] In a specific embodiment, within a set interval, randomly change the random function of the initial road model. According to the specific random function, use the initial road model to generate initial road surface point cloud data with concave and convex feature information having random amplitudes, ranges, and distribution positions.

[0055] Refer to Figure 2 , Figure 2 which is a flowchart of the specific method of step S102 provided by an embodiment of the present application. In some embodiments of the present application, step S102 includes but is not limited to steps S201 to S203. The following combines Figure 2 to introduce these three steps in detail.

[0056] Step S201: Use the initial road model to randomly generate multiple random road surface point cloud data with corresponding amplitudes, ranges, and distribution positions of concave and convex feature information.

[0057] Step S202: According to the concave and convex feature information of the random road surface point cloud data, perform category annotation on the random road surface point cloud data to obtain the randomly road surface point cloud data after category annotation.

[0058] Step S203: Adopt the point cloud three-dimensional rigid body translation method to splice the randomly road surface point cloud data after category annotation into point cloud data of a specified size to obtain the initial road surface point cloud data.

[0059] In a specific embodiment, within a set interval, randomly change the random function of the initial road model. According to the specific random function, use the initial road model to generate random road surface point cloud data with concave and convex feature information having random amplitudes, ranges, and distribution positions. According to the amplitude of the concave and convex feature information of the random road surface point cloud data, perform annotation on each random road surface point cloud data, and adopt the point cloud three-dimensional rigid body translation method to splice and obtain the initial road surface point cloud data with a specified width and length.

[0060] Among them, the annotation of each piece of random road surface point cloud data can be classified into several categories such as flat road surface, small bump, relatively small bump, large bump, very large bump, small depression, relatively small depression, large depression, and very large depression. For example, the random road surface point cloud data with a height ranging from -0.05 to 0.05 can be annotated as a flat road surface, the random road surface point cloud data with a height ranging from 0.05 to 0.15 can be annotated as a small bump, the random road surface point cloud data with a height ranging from 0.15 to 0.25 can be annotated as a relatively small bump, the random road surface point cloud data with a height ranging from 0.25 to 0.35 can be annotated as a large bump, the random road surface point cloud data with a height exceeding 0.35 can be annotated as a very large bump, the random road surface point cloud data with a height ranging from -0.15 to -0.05 can be annotated as a small depression, the random road surface point cloud data with a height ranging from -0.25 to -0.15 can be annotated as a relatively small depression, the random road surface point cloud data with a height ranging from -0.35 to -0.25 can be annotated as a large depression, and the random road surface point cloud data with a height less than -0.35 can be annotated as a very large depression.

[0061] Step S103: Classify and cluster the initial road surface point cloud data according to the convexity and concavity feature information of the initial road surface point cloud data to obtain multiple point cloud clusters of interest.

[0062] Among them, a point cloud cluster of interest contains multiple pieces of point cloud data of interest of the same category. The point cloud data of interest is the initial road surface point cloud data with the convexity and concavity feature information of interest, and the convexity and concavity feature corresponding to the convexity and concavity feature information of interest reaches a preset interference intensity threshold for driving.

[0063] It can be understood that the convexity and concavity feature information of the initial road surface point cloud data describes the corresponding convexity and concavity features. The interference intensities of the convexity and concavity features on driving are different. The point cloud data of interest is a part of the initial road surface point cloud data. The convexity and concavity feature information of the point cloud data of interest is the convexity and concavity feature information of interest. The convexity and concavity feature corresponding to the convexity and concavity feature information of interest reaches a preset interference intensity threshold for driving. That is to say, the point cloud data of interest is the initial road surface point cloud data that has a significant impact on driving.

[0064] In specific implementation, identify the convexity and concavity feature information of the initial road surface point cloud data. According to the recognition result of the convexity and concavity feature information, assign a corresponding category label to each piece of initial road surface point cloud data to determine the category of each piece of initial road surface point cloud data. Then, determine whether the convexity and concavity feature corresponding to the convexity and concavity feature information reaches a preset interference intensity threshold for driving, screen out multiple pieces of point cloud data of interest, and then cluster each piece of point cloud data of interest according to the category label of each piece of point cloud data of interest to aggregate the point cloud data of interest with the same category label to obtain multiple point cloud clusters of interest.

[0065] Refer to Figure 3 , Figure 3It is a flowchart of the specific method of step S103 provided by an embodiment of the present application. In some embodiments of the present application, step S103 includes but is not limited to steps S301 to S303. The following will combine Figure 3 to introduce these three steps in detail.

[0066] Step S301: Input the initial road surface point cloud data into the concave-convex feature semantic segmentation model to predict the concave-convex feature categories of the initial road surface point cloud data, and obtain the point cloud data category prediction result.

[0067] The concave-convex feature semantic segmentation model is trained by a neural network model based on the PointNet++ semantic segmentation network. The point cloud data category prediction result is obtained by using the pre-trained concave-convex feature semantic segmentation model to predict the concave-convex feature categories of the initial road surface point cloud data. In specific implementation, the initial road surface point cloud data is input into the concave-convex feature semantic segmentation model. In the concave-convex feature semantic segmentation model, the initial road surface point cloud data is sampled into multiple point subsets and the concave-convex features of the initial road surface point cloud data are extracted. Then, the Farthest Point Sampling (FPS) algorithm is used to complete the selection of the center point set of the concave-convex features of the initial road surface point cloud data to obtain the corresponding concave-convex feature vectors. Then, the features of the concave-convex feature vectors are extracted, and the extracted features are concatenated to obtain the concave-convex feature concatenated vector. Finally, the concave-convex feature concatenated vector is feature-classified through the corresponding classification layer to predict the concave-convex feature categories of the initial road surface point cloud data, and the point cloud data category prediction result is obtained.

[0068] Step S302: Select the point cloud data from the initial road surface point cloud data according to the point cloud data category prediction result to obtain multiple point clouds of interest.

[0069] In specific implementation, according to the point cloud data category prediction result, then determine whether the concave-convex feature corresponding to the concave-convex feature information has an interference intensity on driving reaching the preset interference intensity threshold. If the interference intensity on driving reaches the preset interference intensity threshold, the initial road surface point cloud data is screened out as the point cloud of interest. After traversing each initial road surface point cloud data, multiple point clouds of interest are obtained.

[0070] Step S303: Cluster each point cloud of interest to obtain multiple clusters of point clouds of interest.

[0071] In specific implementation, according to the categories of the point clouds of interest, the density-based clustering algorithm is used to cluster the point clouds of interest with the same category. The point clouds of interest with the same category and the relative distance meeting the clustering distance condition are clustered into a cluster of point clouds of interest. After traversing each point cloud of interest, multiple clusters of point clouds of interest are obtained.

[0072] Refer to Figure 4 , Figure 4 which is a flowchart of the method before step S301 provided by an embodiment of the present application. In some embodiments of the present application, the method includes but is not limited to steps S401 to S403. The following will introduce these three steps in detail in combination with Figure 4 .

[0073] Step S401: Input the initial road surface point cloud data into the neural network model to predict the concavity and convexity feature categories of the initial road surface point cloud data, and obtain the point cloud data category prediction samples.

[0074] Step S402: Determine the model loss information of the neural network model according to the true category of the initial road surface point cloud data and the point cloud data category prediction samples.

[0075] Step S403: Train the neural network model according to the model loss information. When the neural network model meets the training end condition, obtain the concavity and convexity feature semantic segmentation model.

[0076] In specific implementation, a part of the initial road surface point cloud data is used as the training point cloud data, and another part of the initial road surface point cloud data is used as the verification point cloud data. Then, the training point cloud data is input into the neural network model to train the neural network model. The neural network model is used to predict the concavity and convexity feature categories of the initial road surface point cloud data, and the model loss information of the neural network model is calculated by using the predicted point cloud data category prediction samples and the true category of the initial road surface point cloud data. It is determined whether the neural network model meets the training end condition according to the model loss information of the neural network model, and the weight parameters of the neural network model are updated by using the gradient descent method. Specifically, the gradient of the model loss information with respect to the weight parameters of the neural network model is calculated, and according to the preset learning rate and the calculated gradient, the weight parameters of the neural network model are iteratively updated until the model loss information meets the loss condition (the loss value no longer significantly decreases) or the number of iterations reaches the threshold number, and then the training is ended to obtain the concavity and convexity feature semantic segmentation model.

[0077] Step S104: Fit each point cloud cluster of interest into a form represented by multiple two-dimensional Gaussian functions to obtain an unstructured road concavity and convexity feature model.

[0078] In specific implementation, through the fitting method, each point cloud cluster of interest is respectively fitted with a two-dimensional Gaussian function to obtain parameters such as the amplitude, the center coordinates of the amplitude point, the decay coefficient, and the distribution range of each point cloud cluster of interest. The parameters of the point cloud cluster of interest obtained by fitting are used for model reconstruction to obtain an unstructured road concavity and convexity feature model.

[0079] In a specific embodiment, step S104 specifically includes: representing the point cloud clusters of interest by using a plurality of two-dimensional Gaussian functions, and adopting the least squares method to fit the amplitude, the center coordinates of the amplitude points, the decay coefficient, and the distribution range of the point cloud clusters of interest, so as to obtain an unstructured road concave-convex feature model.

[0080] In a specific embodiment, the expression of the unstructured road concave-convex feature model is:

[0081]

[0082] where Z j (x, y) represents the height of the road surface, and any (x, y) coordinate corresponds to a height value. A j is the amplitude of the point cloud clusters of interest, μ xj is the x-axis coordinate of the amplitude point of the point cloud clusters of interest, μ yj is the y-axis coordinate of the amplitude point of the point cloud clusters of interest, c j is the decay coefficient, is the distribution range of the point cloud clusters of interest in the positive x-axis direction, is the distribution range of the point cloud clusters of interest in the negative x-axis direction, is the distribution range of the point cloud clusters of interest in the positive y-axis direction, is the distribution range of the point cloud clusters of interest in the negative y-axis direction, j = 1, 2,..., m, and m is the total number of the point cloud clusters of interest.

[0083] Please refer to Figure 5 , the embodiment of the present application further provides an unstructured road concave-convex feature model construction device, which can implement the above unstructured road concave-convex feature model construction method. The device includes:

[0084] A first module 501, configured to construct an initial road model based on a Gaussian function;

[0085] A second module 502, configured to randomly generate a plurality of initial road surface point cloud data based on the initial road model; the initial road surface point cloud data includes randomly generated concave-convex feature information;

[0086] A third module 503, configured to classify and cluster the initial road surface point cloud data according to the concave-convex feature information of the initial road surface point cloud data, so as to obtain a plurality of point cloud clusters of interest; the point cloud clusters of interest include a plurality of point cloud data of the same category, the point cloud data of interest is the initial road surface point cloud data with the concave-convex feature information of interest, and the concave-convex feature corresponding to the concave-convex feature information of interest reaches a preset interference intensity threshold for driving;

[0087] The fourth module 504 is configured to fit each point cloud cluster of interest into a form represented by multiple two-dimensional Gaussian functions, thereby obtaining an unstructured road bump feature model.

[0088] The specific implementation manner of the unstructured road bump feature model construction device is basically the same as the specific embodiments of the above-mentioned unstructured road bump feature model construction method, and will not be elaborated herein.

[0089] Figure 6 It is a block diagram of an electronic device shown according to an exemplary embodiment.

[0090] Next, reference will be made to Figure 6 to describe the electronic device 600 according to this embodiment of the present disclosure. Figure 6 The electronic device 600 shown is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present disclosure.

[0091] As Figure 6 shown, the electronic device 600 is presented in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0092] Among them, the storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 executes the steps according to various exemplary embodiments of the present disclosure described in the above-mentioned unstructured road bump feature model construction method section of this specification.

[0093] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 6201 and / or a cache storage unit 6202, and may further include a read-only storage unit (ROM) 6203.

[0094] The storage unit 620 may also include a program / utilities 6204 having a set (at least one) of program modules 6205. Such program modules 6205 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. The implementation of a network environment may be included in each or some combination of these examples.

[0095] The bus 630 may represent one or more of several types of bus structures, including a storage unit bus or a storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any bus structure in a variety of bus structures.

[0096] The electronic device 600 can also communicate with one or more external devices 600' (such as a keyboard, a pointing device, a Bluetooth device, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (such as a router, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 650. Moreover, the electronic device 600 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 660. The network adapter 660 can communicate with other modules of the electronic device 600 through the bus 630. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0097] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the above-mentioned method for constructing an unstructured road bump feature model is implemented.

[0098] The method for constructing an unstructured road bump feature model and related devices provided by the embodiments of the present application utilize an initial road model constructed based on a Gaussian function to randomly generate a plurality of initial road surface point cloud data containing random bump feature information. According to the bump feature information of the initial road surface point cloud data, the initial road surface point cloud data is classified and clustered to obtain a plurality of point cloud clusters of interest points composed of point cloud data of interest that have a significant impact on driving. Then, each point cloud cluster of interest is fitted to be represented in the form of a plurality of two-dimensional Gaussian functions to obtain an unstructured road bump feature model. Since a plurality of initial road surface point cloud data are randomly generated using the initial road model, and then the point cloud data of interest that have a significant impact on driving are selected from them, and the unstructured road bump feature model is reconstructed using the point cloud clusters of interest composed of point cloud data of the same category, the bump features of the unstructured road surface can be accurately expressed, and the perception accuracy of the road surface can be improved.

[0099] Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above methods according to the embodiments of the present disclosure.

[0100] The program product can adopt any combination of one or more readable media. The readable media can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, but not be limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0101] The computer-readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries the readable program code. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The readable storage medium can also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program used by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium can be transmitted by any appropriate medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination of the above.

[0102] The program code for performing the operations of the present disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, executed as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., by using an Internet service provider to connect through the Internet).

[0103] Those skilled in the art can understand that the above-mentioned modules can be distributed in the device according to the description of the embodiments, or can be correspondingly changed and distributed in one or more devices that are only different from this embodiment. The modules of the above embodiments can be combined into one module, or can be further split into multiple sub-modules.

[0104] The exemplary embodiments of the present disclosure have been specifically shown and described above. It should be understood that the present disclosure is not limited to the detailed structures, settings or implementation methods described herein; on the contrary, the present disclosure is intended to cover various modifications and equivalent settings included within the spirit and scope of the appended claims.

Claims

1. A method for constructing an unstructured road bump feature model, characterized in that, Including: Construct an initial road model based on the Gaussian function; Based on the initial road model, randomly generate a plurality of initial pavement point cloud data; The initial pavement point cloud data includes randomly generated bump feature information; According to the bump feature information of the initial pavement point cloud data, classify and cluster the initial pavement point cloud data to obtain a plurality of point cloud clusters of interest; the point cloud clusters of interest include a plurality of point cloud data of the same category, and the point cloud data of interest is the initial pavement point cloud data with the bump feature information of interest, and the bump feature corresponding to the bump feature information of interest has an interference intensity on driving reaching a preset interference intensity threshold; Fit each of the point cloud clusters of interest into a form represented by a plurality of two-dimensional Gaussian functions to obtain an unstructured road bump feature model; The expression of the unstructured road bump feature model is: ， Among them, represents the height of the road surface, and any (x, y) coordinate corresponds to a height value. A j is the amplitude of the point cloud cluster of interest, μ xj is the x-axis coordinate of the amplitude point of the point cloud cluster of interest, μ yj is the y-axis coordinate of the amplitude point of the point cloud cluster of interest, c j is the descent coefficient, is the distribution range of the point cloud cluster of interest in the positive x-axis direction, is the distribution range of the point cloud cluster of interest in the negative x-axis direction, is the distribution range of the point cloud cluster of interest in the positive y-axis direction, is the distribution range of the point cloud cluster of interest in the negative y-axis direction, j = 1, 2,..., m, where m is the total number of point cloud clusters of interest.

2. The method for constructing an unstructured road bump feature model according to claim 1, wherein The expression of the initial road model is: ， Among them, The function represents the position of the road surface point cloud in the y-axis direction at the position, 、 and are the x, y, and z coordinates of the point cloud respectively, and A i , B i and C i are all Gaussian function parameter values, ω i (x) is a uniformly distributed random function, i = 1, 2,..., n, and n is the total number of random road surface point cloud data.

3. The method for constructing an unstructured road bump feature model according to claim 1, wherein The step of randomly generating a plurality of initial pavement point cloud data based on the initial road model includes: Using the initial road model, randomly generate a plurality of random pavement point cloud data with bump feature information of corresponding amplitudes, ranges and distribution positions; According to the bump feature information of the random pavement point cloud data, perform category annotation on the random pavement point cloud data to obtain the randomly generated pavement point cloud data after category annotation; Adopt a point cloud three-dimensional rigid body translation method to splice the randomly generated pavement point cloud data after category annotation into point cloud data of a specified size to obtain the initial pavement point cloud data.

4. The method for constructing an unstructured road bump feature model according to claim 1, wherein The step of classifying and clustering the initial pavement point cloud data according to the bump feature information of the initial pavement point cloud data to obtain a plurality of point cloud clusters of interest includes: Input the initial pavement point cloud data into a bump feature semantic segmentation model to perform bump feature category prediction on the initial pavement point cloud data to obtain a point cloud data category prediction result; the bump feature semantic segmentation model is trained by a neural network model based on the PointNet++ semantic segmentation network; According to the point cloud data category prediction result, select the initial pavement point cloud data to obtain a plurality of the point cloud data of interest; Cluster each of the point cloud data of interest to obtain a plurality of the point cloud clusters of interest.

5. The method for constructing an unstructured road bump feature model according to claim 4, characterized in that Before inputting the initial pavement point cloud data into the bump feature semantic segmentation model to perform bump feature category prediction on the initial pavement point cloud data to obtain a point cloud data category prediction result, it further includes: Input the initial pavement point cloud data into the neural network model to perform bump feature category prediction on the initial pavement point cloud data to obtain a point cloud data category prediction sample; According to the true category of the initial pavement point cloud data and the point cloud data category prediction sample, determine the model loss information of the neural network model; According to the model loss information, train the neural network model, and when the neural network model meets the training end condition, obtain the bump feature semantic segmentation model.

6. The method for constructing an unstructured road bump feature model according to claim 1, characterized in that, The step of fitting each of the point cloud clusters of interest into a form represented by a plurality of two-dimensional Gaussian functions to obtain an unstructured road bump feature model includes: The point cloud cluster of interest is represented by multiple two-dimensional Gaussian functions, and the least squares method is used to fit the amplitude, the center coordinates of the amplitude points, the decay coefficient, and the distribution range of the point cloud cluster of interest to obtain the unstructured road unevenness feature model.

7. An unstructured road bump feature model construction device, characterized in that It includes: A first module for constructing an initial road model based on Gaussian functions; A second module for randomly generating multiple initial road surface point cloud data based on the initial road model; The initial road surface point cloud data contains randomly generated unevenness feature information; A third module for classifying and clustering the initial road surface point cloud data according to the unevenness feature information of the initial road surface point cloud data to obtain multiple point cloud clusters of interest; the point cloud clusters of interest contain multiple point cloud data of the same category, and the point cloud data of interest is the initial road surface point cloud data with unevenness feature information of interest, and the unevenness feature corresponding to the unevenness feature information of interest reaches a preset interference intensity threshold for driving; A fourth module for fitting each of the point cloud clusters of interest into a form represented by multiple two-dimensional Gaussian functions to obtain an unstructured road unevenness feature model; The expression of the unstructured road unevenness feature model is: ， Among them, represents the height of the road surface, and any (x, y) coordinate corresponds to a height value, A j is the amplitude of the point cloud cluster of interest, μ xj is the x-axis coordinate of the amplitude point of the point cloud cluster of interest, μ yj is the y-axis coordinate of the amplitude point of the point cloud cluster of interest, c j is the descent coefficient, is the distribution range of the point cloud cluster of interest in the positive x-axis direction, is the distribution range of the point cloud cluster of interest in the negative x-axis direction, is the distribution range of the point cloud cluster of interest in the positive y-axis direction, is the distribution range of the point cloud cluster of interest in the negative y-axis direction, j = 1, 2,..., m, where m is the total number of point cloud clusters of interest.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the unstructured road unevenness feature model construction method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the unstructured road unevenness feature model construction method according to any one of claims 1 to 6.

Images