Road model generation method and device, electronic equipment, medium and program product

Abstract

The present disclosure provides a road model generation method and device, electronic equipment, medium and program product, relates to the technical field of computers, and particularly relates to the fields of automatic driving, map modeling, road modeling, etc. The specific implementation scheme is: acquiring road data corresponding to a base point in a road to be modeled, the road data including attribute parameters of road elements at a position corresponding to the base point; generating target modeling data corresponding to the base point based on a target template matched with the road data, wherein the target template includes a road element model at the position corresponding to the base point, and the target modeling data is used to generate a road model at the position corresponding to the base point; and generating a road model of the road to be modeled based on the target modeling data, wherein the road model is used to generate map data. The present disclosure can improve the efficiency of road modeling.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of computer, and particularly relates to the fields of automatic driving, map modeling, road modeling, etc. Specifically, it relates to a road model generation method and device, an electronic device, a medium and a program product. BACKGROUND

[0002] In the related art, a high-precision map plays an important role in an automatic driving vehicle. At present, a data production mode of the high-precision map is mainly to collect various sensor data of road information through a professional collection vehicle, and to complete generation of basic data in a manner of automatic algorithm and manual labeling, wherein the basic data includes various map elements, such as a road, various elements in the road, a building, etc. SUMMARY

[0003] The present disclosure provides a road model generation method and device, an electronic device, a medium and a program product.

[0004] According to a first aspect of the present disclosure, a road model generation method is provided, comprising:

[0005] obtaining road data corresponding to a base point in a road to be modeled, wherein the road data includes attribute parameters of road elements at a position corresponding to the base point;

[0006] generating target modeling data corresponding to the base point based on the road data and a target template matched with the road data, wherein the target template includes a road element model at the position corresponding to the base point, and the target modeling data is used to generate a road model of the position corresponding to the base point;

[0007] generating a road model of the road to be modeled based on the target modeling data, wherein the road model is used to generate map data.

[0008] According to a second aspect of the present disclosure, a road model generation device is provided, comprising:

[0009] an obtaining module, configured to obtain road data corresponding to a base point in a road to be modeled, wherein the road data includes attribute parameters of road elements at a position corresponding to the base point;

[0010] a first generating module, configured to generate target modeling data corresponding to the base point based on the road data and a target template matched with the road data, wherein the target template includes a road element model at the position corresponding to the base point, and the target modeling data is used to generate a road model of the position corresponding to the base point;

[0011] A second generation module is configured to generate a road model of the to-be-modeled road based on the target modeling data, and the road model is used to generate map data.

[0012] According to a third aspect of the present disclosure, an electronic device is provided, comprising:

[0013] at least one processor; and

[0014] a memory in communication with the at least one processor; wherein

[0015] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of the first aspect.

[0016] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to perform the method of the first aspect.

[0017] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the method of the first aspect.

[0018] In the embodiments of the present disclosure, the target modeling data corresponding to the position of the base point is generated by using the road data and the target template, so that the road model of the position corresponding to the base point can be completed according to the target modeling data. Compared with the road modeling means in the related art, various target templates can be reused, which is beneficial to improve the efficiency of road modeling, and at the same time, is beneficial to reduce the cost of road modeling. BRIEF DESCRIPTION OF DRAWINGS

[0019] The accompanying drawings serve to better understand the present scheme and do not constitute a limitation on the present disclosure. Among them:

[0020] Figure 1 is a flowchart of a road model generation method provided by an embodiment of the present disclosure;

[0021] Figure 2 is a schematic diagram of a road template provided by an embodiment of the present disclosure;

[0022] Figure 3 is a schematic diagram of the first road image sliding relative to the second road image in an embodiment of the present disclosure;

[0023] Figure 4 is a schematic diagram of the score calculated in an embodiment of the present disclosure;

[0024] Figure 5 is a structural schematic diagram of a network model in an embodiment of the present disclosure;

[0025] Figure 6 is a schematic diagram of a road model generated in an embodiment of the present disclosure;

[0026] Figure 7 is a schematic diagram of a module of a road model generation system provided in an embodiment of the present disclosure;

[0027] Figure 8 is one of the structural schematic diagrams of a road model generation apparatus provided in an embodiment of the present disclosure;

[0028] Figure 9 is the second structural schematic diagram of a road model generation apparatus provided in an embodiment of the present disclosure;

[0029] Figure 10 is a structural schematic diagram of a third generation module provided in an embodiment of the present disclosure;

[0030] Figure 11 is a structural schematic diagram of a calculation module provided in an embodiment of the present disclosure;

[0031] Figure 12 is a structural schematic diagram of an acquisition module provided in an embodiment of the present disclosure;

[0032] Figure 13 is a block diagram of an electronic device for implementing the road model generation method provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0033] Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to assist in understanding, which should be considered in their context only. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Also, in order to be clear and concise, the description below omits the description of well-known functions and structures.

[0034] Please refer to Figure 1 , Figure 1 is a flowchart of a road model generation method provided in an embodiment of the present disclosure, which comprises the following steps:

[0035] Step S101, acquiring road data corresponding to a base point in a road to be modeled, the road data comprising attribute parameters of road elements at a position corresponding to the base point;

[0036] In step S102, target modeling data corresponding to the base point is generated based on the road data and a target template matched with the road data, wherein the target template comprises a road element model in a position corresponding to the base point, and the target modeling data is used to generate a road model of the position corresponding to the base point.

[0037] In step S103, a road model of the road to be modeled is generated based on the target modeling data, and the road model is used to generate map data.

[0038] The road to be modeled can be various types of roads in a high-precision map generation scenario, for example, a one-way road, a two-way road, an intersection road, etc. The base point can refer to one position point in the road to be modeled.

[0039] It can be understood that a plurality of base points can be predetermined in the road to be modeled, and the plurality of base points correspond to different positions in the road to be modeled respectively. Thus, a plurality of target modeling data corresponding to the plurality of base points respectively can be generated based on the method in the embodiments of the present disclosure. Thus, road models of different positions in the road to be modeled can be constructed based on the plurality of target modeling data respectively, thereby completing the modeling process of the road to be modeled.

[0040] The road element can comprise various road elements to be modeled in the road modeling process. For example, the road element can comprise a lane line, a road edge, a guardrail, etc. Correspondingly, the attribute parameter of the road element can comprise a position of a lane line, a road width, a lane width, a distance between a guardrail and a road boundary, etc. Thus, since the target template comprises a model of each road element in the road of the position corresponding to the base point, and the road data comprises attribute parameters of each road element in the road of the position corresponding to the base point, the road model of the position corresponding to the base point can be generated by fusing the target template and the road data.

[0041] The target modeling data can be data obtained by filling attribute parameters in the road data into the target template.

[0042] In an embodiment of the present disclosure, a large number of basic templates can be created in advance, and different basic templates can be created for different types of roads, for example, corresponding basic templates can be created for roads of different lanes. In addition, corresponding basic templates can also be created based on various road positions, for example, the road can be divided into normal traffic positions, lane changing positions, intersection positions, etc., and corresponding basic templates can be created for each position. The target template can be a basic template that matches the road at the position corresponding to the base point among the pre-created basic templates, and the matching of the road at the position corresponding to the base point can refer to the same number of lanes and the same type of lanes. Specifically, when it is necessary to create a road model of the road to be modeled, the target template can be determined from the pre-created basic templates according to the number of lanes and the type of lanes of the road to be modeled.

[0043] The road data can be data in the bounding box data of each element corresponding to the road to be modeled, which is extracted based on the automatic identification process in the related art. Specifically, before the high-precision map is generated, it is usually necessary to collect data in each road element in the road to be modeled based on a collection vehicle, and then generate the bounding box data based on the collected road data. That is, the road data is real road data collected for an actual road. It can be understood that the base point can be a sampling point in the sampling process.

[0044] Since the model of various traffic elements is usually required in the map data generation process, and the road belongs to an important traffic element in the map data generation process, the road model in the map data can be generated based on the method provided in the embodiments of the present disclosure.

[0045] In this implementation, the target modeling data corresponding to the position of the base point is generated by using the road data and the target template. In this way, the road model of the position corresponding to the base point can be completed by following the target modeling data. Compared with the road modeling means in the related art, various target templates can be reused, which is beneficial to improve the efficiency of road modeling and reduce the cost of road modeling.

[0046] Optionally, before the target modeling data corresponding to the base point is generated based on the road data and the target template matching the road data, the method further comprises:

[0047] Based on the road data and the basic templates in the template library, a plurality of initial modeling data matching the road data are generated, wherein the road elements corresponding to the basic templates match the road elements corresponding to the road data, and the initial modeling data are used to generate the road model of the position corresponding to the base point;

[0048] Target score values ​​are calculated for each of the initial modeling data based on the road data, and the target score values ​​are used to characterize the degree of matching between the road data and the base template;

[0049] The basic template corresponding to the initial modeling data with the highest target score among the multiple basic templates is determined as the target template corresponding to the road data.

[0050] The aforementioned template library can be a collection of pre-built templates from a large number of different base templates. For example, please see... Figure 2 In one embodiment of this disclosure, the template library may include one-way road templates, two-way road templates, and intersection templates. The two-way road templates consist of n one-way road templates, and the intersection templates consist of n two-way road templates, where n is an integer greater than or equal to 2. Specifically, each basic template includes three parts: type, structure, and scalar values. For example, the type of a one-way road template may include ordinary roads and lane-changing zones. Correspondingly, the structure of a one-way road template may include the number of lanes and the number of boundaries. The scalar values ​​of a one-way road template may include the width of each lane and the distance between the road boundary and the nearest lane line. Thus, after determining the target template corresponding to the road data, it is only necessary to identify the above-mentioned scalar values ​​in the road to be modeled and fill the specific values ​​of the scalar values ​​into the target template to obtain the target modeling data. That is, in one embodiment of this disclosure, the road elements may include the above-mentioned type and structure, and the attribute parameters of the road elements may include the scalar values.

[0051] The matching of road elements corresponding to the aforementioned basic template with road elements in the road to be modeled can mean that the target road element corresponding to the basic template matches the target road element in the road to be modeled. The target road element can be a subset of road elements in the road to be modeled; for example, the target road element can include the aforementioned types and structures. Since the road to be modeled includes other road elements besides the target road element, such as boundary types (e.g., in related technologies, road boundaries can use guardrails, double yellow lines, single solid lines, etc.), and the basic template can also include other road elements besides the target road element, different basic templates can be generated based on different other road elements when the target road element is the same, multiple basic templates matching the target road element in the template library may be found. In this case, a set of initial modeling data can be generated based on each matched basic template and the road data, thus obtaining the multiple initial modeling data sets.

[0052] Then, the matching degree between the base template and the road data in each initial modeling data set can be evaluated separately, and a target score value corresponding to each initial modeling data set can be obtained. The higher the target score value, the higher the matching degree between the base template and the road data. Thus, the base template corresponding to the initial modeling data with the highest target score value among the multiple base templates can be determined as the target template corresponding to the road data. This is beneficial for improving the matching degree between the subsequently constructed road model and the road to be modeled.

[0053] The base template corresponding to the initial modeling data with the highest target score mentioned above refers to the base template for generating the initial modeling data with the highest target score.

[0054] In this implementation, multiple initial modeling data are generated based on a base template and road data, and each initial modeling data is scored. Then, the base template corresponding to the initial modeling data with the highest score among the multiple initial modeling data is determined as the target template. This helps to improve the matching degree between the road model constructed subsequently and the road to be modeled.

[0055] Optionally, the step of generating multiple initial modeling data matching the road data based on the road data and basic templates in the template library includes:

[0056] Identify at least two different base templates in the template library that match the road data;

[0057] At least two initial modeling data that match the road data are generated based on each basic template, wherein each basic template corresponds to at least two initial modeling data.

[0058] The multiple initial modeling data that match the road data include: at least two initial modeling data corresponding to the at least two different basic templates.

[0059] Specifically, since road data is obtained by processing data, and since both the data adoption process and the data processing process may introduce errors, the road data can be deformed based on the possible errors. Initial modeling data can then be generated based on the road data before and after the deformation, along with each basic template, thereby obtaining at least two initial modeling data corresponding to each basic template.

[0060] In this implementation, at least two initial modeling data are generated based on each basic template. This increases the number of optional initial modeling data, thereby improving the effectiveness of the determined target template and, consequently, improving the matching degree between the subsequently constructed road model and the road to be modeled.

[0061] Optionally, generating at least two initial modeling data matching the road data based on each of the basic templates includes:

[0062] The target parameters in the road data are scaled to obtain at least two first target road data, the at least two first target road data including: the road data and the road data after scaling the target parameters;

[0063] For each of the at least two basic templates, at least two initial modeling data corresponding to the basic template are generated based on the basic template and the at least two first target road data.

[0064] The target parameter mentioned above can be an attribute parameter such as road width or lane width. The following explanation uses lane width as an example to further illustrate the method provided in this embodiment.

[0065] Assuming the lane width error range may be [-d, d], then m error values ​​can be sampled between [-d, d] at intervals of s, where the value of d can be selected based on practical experience. Specifically, it can be represented as: [a1, a2…a ... mThen, the lane width of the road data is added to each error value to obtain m scaled road data, where m can be an integer greater than or equal to 2. Thus, the m scaled road data and the original road data together form m+1 first target road data. Then, each first target road data is used with the base template to generate a chef modeling data, resulting in m+1 initial modeling data corresponding to each base template. Specifically, when the road width error value is a... i The corresponding scaled road width can be expressed as:

[0066] list[Instance_lane_width=lane_width+a i ]

[0067] In this implementation, the target parameters in the road data are scaled, and the initial modeling data is formed by combining the scaled road data with the basic template. In this way, the errors in the sampling and processing of the road data can be taken into account during the modeling and matching process, which is conducive to improving the matching degree between the subsequently constructed road model and the road to be modeled.

[0068] Optionally, the plurality of initial modeling data includes first modeling data, which is initial modeling data generated based on the road data and a first basic template, wherein the first basic template is a basic template in the template library, and the step of calculating the target score value for each of the initial modeling data based on the road data includes:

[0069] A first road image is determined based on the road data, and a second road image is determined based on the first base template;

[0070] When the first road image and the second road image at least partially overlap, the target score value corresponding to the first modeling data is calculated based on the distance between each road element in the first road image and the corresponding road element in the second road image.

[0071] Specifically, since the road data includes attribute parameters of each element in the road to be identified, the basic image content of the road to be identified can be reconstructed based on the road data. For example, please refer to [link to relevant documentation]. Figure 3This is a first road image reconstructed from the road data, based on the positions of each lane line (implied arrow) and boundary line (dashed arrow) in the road data. The first base template includes a pre-created road model of a template road. The first base template can be pre-configured with initial attribute parameters for each road element. When generating initial modeling data based on the first base template and the road data, the attribute parameters from the road data can replace the initial attribute parameters in the first base template. Therefore, a second road image corresponding to the template road can be reconstructed based on the first base template.

[0072] The higher the degree of overlap between the positions of each road element in the first road image and each road image in the second road image, the higher the degree of proximity between the first road image and the second road image. In other words, the higher the degree of matching between the road data and the first basic template, the higher the target score value of the first modeling data.

[0073] Based on this, in this embodiment of the disclosure, the target score can be calculated using the following formula:

[0074]

[0075]

[0076] Wherein, match_score represents the target score value, and w i Let P(Δdis) represent the weight value corresponding to the i-th road element, λ represent the pre-configured coefficients, and Δdis represent the distance between the i-th road element in the first road image and the i-th road element in the second road image. ∑w i P(Δdis) represents the sum of the scores of all road elements, ∑w i This represents the sum of the weight values ​​corresponding to all road elements.

[0077] As shown in formula (2), the smaller Δdis is, the larger the corresponding P(Δdis) is, and correspondingly, the larger the match_score will be. That is, the match_score can represent the sum of distances between all road elements in the first road image and the corresponding road elements in the second road image. Thus, the match_score can be used to represent the degree of matching between the road data and the base template.

[0078] In this embodiment, the target score value corresponding to the first modeling data is calculated based on the distance between each road element in the first road image and the corresponding road element in the second road image, thereby realizing the calculation process of the matching degree between the road data and each basic template.

[0079] Optionally, when the first road image and the second road image at least partially overlap, calculating the target score value corresponding to the first modeling data based on the distance between each road element in the first road image and the corresponding road element in the second road image includes:

[0080] Determine at least two distinct target relative positions, wherein, when the first road image and the second road image are at the target relative positions, at least one road element in the first road image and the second road image has an overlapping position;

[0081] Calculate the initial score value corresponding to the relative position of each target to obtain at least two initial score values;

[0082] The initial score with the highest score among the at least two initial score values ​​is determined as the target score value corresponding to the first modeling data.

[0083] Since the road elements in the first road image and their corresponding road elements in the second road image may not completely overlap, and the image sizes of the first and second road images may also differ, the process of calculating the target score value between the first and second road images can be achieved by sliding the first and second road images relative to each other to find the relative position with the highest degree of overlap between their road elements. The initial score calculated based on this relative position is then determined as the target score value corresponding to the first modeling data.

[0084] Specifically, please see Figure 3 The method involves sliding a first road image relative to a second road image, calculating initial scores at multiple different locations during the sliding process, and then determining the highest initial score as the target score. This facilitates the calculation of the score at the relative location where the road elements in the first and second road images have the highest overlap.

[0085] In the first road image and the second road image mentioned above, the existence of at least one road element with overlapping positions means that road elements of the same type overlap in position. For example, the lane lines in the first road image and the lane lines in the second road image overlap in position, or the boundary lines in the first road image and the boundary lines in the second road image overlap in position.

[0086] Please see Figure 3 The first road image can be slid relative to the second road image from a first relative position to a second relative position, and during the sliding process, at least a target relative position is determined. The first relative position is the location where the left boundary line of the first road image coincides with the right boundary line of the second road image, and the second relative position is the location where the right boundary line of the first road image coincides with the left boundary line of the second road image. During this sliding process, all locations where at least one road element overlaps between the first and second road images can be found. For example, see [link to relevant documentation]. Figure 3 When the first road image slides to the third relative position, the middle lane line in the first road image coincides with the middle lane line in the second road image, and the third relative position can be determined as a target relative position. It is understood that the above-mentioned at least two different target relative positions include the first relative position and the second relative position.

[0087] Please see Figure 4 In one embodiment of this disclosure, the calculation result of the score value corresponding to the first basic template is obtained, wherein after scaling the target parameters in the road data, three first target road data (corresponding to) are obtained. Figure 4 The three dimensions in the image), and simultaneously, during the sliding process of the first road image, there are three relative positions of the target (corresponding to...). Figure 4 (The three positions in the text) Thus, for the first basic template, nine rating values ​​can be calculated, and the highest value of these nine rating values, 0.9, is determined as the rating value corresponding to the first basic template. In this way, the rating value of each basic template can be calculated, and the basic template with the highest rating value is determined as the target template.

[0088] In this embodiment, since the number of relative target positions is relatively limited during the sliding of the first road image, the computational load in the target score calculation process is reduced. Simultaneously, the relative position with the highest overlap between road elements in the first and second road images is typically one of the aforementioned at least two different target relative positions, or a position close to one of the aforementioned at least two different target relative positions. This helps improve the accuracy of the calculated target score.

[0089] Optionally, obtaining road data corresponding to base points in the road to be modeled includes:

[0090] A first grayscale image and a trajectory mask image are generated based on an initial image, wherein the initial image is an image containing the road to be modeled, the first grayscale image is a grayscale image of the initial image, and the trajectory mask image includes the trajectory information of the road to be modeled;

[0091] The first grayscale image and the trajectory mask image are input into a pre-trained network model for semantic recognition to obtain multiple semantic images output by the network model, wherein the semantic images include attribute parameters of one dimension of the base point;

[0092] The road data is generated based on the multiple semantic images.

[0093] The aforementioned network model can be a common model used in related technologies for extracting road semantics, such as a model trained based on deeplabv3+ or Unet. It is understood that the network model can perform semantic extraction on the road to be modeled based on the first grayscale image and the trajectory mask image to obtain the aforementioned road data. For example, please refer to... Figure 5 The network model includes an encoder, a decoder, and a perception layer. After the first grayscale image and the trajectory mask image are input into the network model, they are processed sequentially by the encoder, decoder, and perception layer to obtain the following road inputs: road semantic image, trajectory direction semantic image, road width semantic image, lane number semantic image, and distance semantic image between the road boundary and the nearest lane line.

[0094] The aforementioned first grayscale image adopts an arbitrary form of Bird's Eye View (BEV) image, for example, it can be a single-channel BEV grayscale image. The aforementioned trajectory mask image can include any point on the road to be modeled, representing the road direction. The generation process of the trajectory mask image can be as follows:

[0095] First, a kd-demension tree (KDTree) spatial index is used to locate all trajectory information within the BEV map area. Then, the trajectories are encoded using a three-layer encoding. The first layer represents the scope of the trajectory information: 1 for areas with trajectories and 0 for areas without trajectories. The second layer represents the vertical direction information of the trajectory information: 1 for north ±90 degrees and 0 otherwise. The third layer represents the horizontal direction information of the trajectory information: 1 for east ±90 degrees and 0 otherwise. Together, these three layers of encoding express whether trajectory information exists at the current location and the quadrant to which the corresponding trajectory orientation belongs. This encoding method provides sufficient information while improving system robustness.

[0096] The specific inputs and outputs of the above network model can include:

[0097] The road semantic mask image has pixel values ​​in the range [0,1], indicating whether the location corresponding to the pixel is within the valid road area.

[0098] The road width mask image has pixel values ​​> 0. To facilitate training, this scheme uses a discretization method with a precision of 10cm. The discretization formula is as follows:

[0099] y = int(road_width * 10)

[0100] The lane number mask image, with pixel values ​​> 0, represents the actual number of lane lines on the road at the given location.

[0101] The distance from the road boundary to the mask image, with pixel values ​​ranging from >=0, represents the vertical distance from that location to the nearest road boundary. Discretized using a 10cm precision method, the discretization formula is as follows:

[0102] y = int(boundary_dis * 10)

[0103] The distance mask image for oncoming roads, with pixel values ​​ranging from 0 to 0, represents the vertical distance from the location to the boundary of the nearest oncoming road. It is discretized with a precision of 10cm, and the discretization formula is as follows:

[0104] y = int(gap_width * 10)

[0105] It is understood that the aforementioned road data may include not only the data output by the network model, but also data generated based on the data output by the network model, such as lane width (lane_width). Specifically, since the road width (road_width) of the road to be modeled can be determined from the aforementioned road width mask image, and the number of lanes (lane_num) of the road to be modeled can be determined from the aforementioned lane number mask image, the lane width (lane_width) of the road to be modeled can be calculated based on the following formula:

[0106]

[0107] In this embodiment, a pre-trained network model is used to perform semantic recognition on the first grayscale image and the trajectory mask image to obtain the road data of the road to be modeled, thereby achieving automated acquisition of road data.

[0108] Optionally, the plurality of semantic images includes directional semantic images, which include the initial direction of each pixel in the road to be modeled. Generating the road data based on the plurality of semantic images includes:

[0109] Based on the initial directions of each pixel in the directional semantic image that is within a preset range from the base point, the initial direction of the base point is corrected to obtain the target direction of the base point, wherein the road data includes the target direction.

[0110] In one embodiment of this disclosure, the pixel value range of the aforementioned directional semantic image is [0, 36], corresponding to a clockwise rotation of angle / 10 with true north as the reference. For example, when the directional angle of a pixel is 40°, the pixel value of that pixel is 40 / 10 = 4. Thus, various directions can be covered. The pixel value of each pixel in the directional semantic image is used to characterize the initial direction of that pixel.

[0111] Since the accuracy of directional information in directional semantic images is not 100%, gaps may occur, meaning that a pixel at a certain location is 0, while pixels within a certain radius around it are non-zero. Therefore, this disclosure designs a weighted mask voting method to handle the problem caused by such outliers. The specific steps are as follows: based on a base point, find all pixels in a circular region of radius d surrounding it, and calculate the weighted average pixel value according to the following formula, which is used as the target direction after correction for that base point. The specific correction formula is:

[0112]

[0113] Wherein, the v avg This represents the corrected pixel value. A total of n pixels were found around the base point, and v... i Let be the pixel value of the i-th pixel. λ is a preset coefficient, and dis is the distance between the i-th pixel and the base point.

[0114] In this embodiment, the initial direction of the base point is corrected based on the initial direction of each pixel in the directional semantic image that is within a preset range from the base point, thereby avoiding the problem of holes in the base point and thus improving the accuracy of the acquired road information.

[0115] Optionally, the road to be modeled includes multiple base points, and the multiple base points are multiple location points arranged at equal intervals along the extension direction of the road to be modeled. The location corresponding to the first base point is the road segment between the first base point and the second base point. The first base point is any base point among the multiple base points. The second base point is the base point among the multiple base points that is adjacent to the first base point.

[0116] In one embodiment of this disclosure, along the lane direction of the road to be modeled, the second base point may be the next base point after the first base point.

[0117] Specifically, the target modeling data for each base point can be determined based on the method described in the above embodiments, so as to obtain multiple target modeling data corresponding one-to-one with the multiple base points. For example... Figure 6 As shown, a road model for each base point can be constructed based on the multiple target modeling data, thereby obtaining the lane model of the road to be modeled.

[0118] In this embodiment, by determining multiple equally spaced base points in the road to be modeled, the road modeling process of the road to be modeled can be completed simply by calculating the target modeling data of each base point.

[0119] Please see Figure 7 This is a schematic flowchart of a road model generation system provided in an embodiment of the present disclosure. The road model generation system includes:

[0120] Manually create modules to create the basic templates in the above embodiments;

[0121] The deep learning module is used to train the network model described in the above embodiments;

[0122] The generation module is used to match N instances based on the basic template generated by the manual creation module and the road information identified by the deep learning module, wherein the instances are the initial modeling data described in the above embodiments;

[0123] The pre / post-processing module is used to generate lane line instances based on the road model, then calculate the score values ​​of the N instances to obtain the target template, and then complete the road modeling process based on the lane line instances, the target template and the road data.

[0124] The specific implementation process in this embodiment is similar to that in the above embodiments, and can achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0125] Please see Figure 8 This is a schematic diagram of the structure of a road model generation device 800 provided in an embodiment of the present disclosure. The road model generation device 800 includes:

[0126] The acquisition module 801 is used to acquire road data corresponding to the base point in the road to be modeled, wherein the road data includes attribute parameters of the road element at the location corresponding to the base point;

[0127] The first generation module 802 is used to generate target modeling data corresponding to the base point based on the road data and the target template that matches the road data. The target template includes road element models at the location corresponding to the base point, and the target modeling data is used to generate a road model at the location corresponding to the base point.

[0128] The second generation module 803 is used to generate a road model of the road to be modeled based on the target modeling data, and the road model is used to generate map data.

[0129] Optionally, the device further includes:

[0130] The third generation module 804 is used to generate multiple initial modeling data that match the road data based on the road data and the basic template in the template library. The road elements corresponding to the basic template match the road elements in the road to be modeled. The initial modeling data is used to generate a road model at the location corresponding to the base point.

[0131] Calculation module 805 is used to calculate a target score value for each of the initial modeling data based on the road data, wherein the target score value is used to characterize the degree of matching between the road data and the basic template;

[0132] The determining module 806 is used to determine the basic template corresponding to the initial modeling data with the highest target score among the multiple basic templates as the target template corresponding to the road data.

[0133] Optionally, the third generation module 804 includes:

[0134] The first determining submodule 8041 is used to determine at least two different basic templates in the template library that match the road data;

[0135] The first generation submodule 8042 is used to generate at least two initial modeling data that match the road data based on each basic template, wherein each basic template corresponds to at least two initial modeling data.

[0136] The multiple initial modeling data that match the road data include: at least two initial modeling data corresponding to the at least two different basic templates.

[0137] Optionally, the first generation submodule 8042 is used to scale the target parameters in the road data to obtain at least two first target road data, wherein the at least two first target road data include: the road data and the road data after scaling the target parameters;

[0138] The first generation submodule 8042 is further configured to generate at least two initial modeling data corresponding to each of the at least two basic templates, based on the basic template and the at least two first target road data.

[0139] Optionally, the plurality of initial modeling data includes first modeling data, which is initial modeling data generated based on the road data and the first basic template, wherein the first basic template is a basic template in the template library, and the calculation module 805 includes:

[0140] The second determining submodule 8051 is used to determine a first road image based on the road data, and to determine a second road image based on the first base template;

[0141] The calculation submodule 8052 is used to calculate the target score value corresponding to the first modeling data based on the distance between each road element in the first road image and the corresponding road element in the second road image when the first road image and the second road image at least partially overlap.

[0142] Optionally, the calculation submodule 8052 is used to determine at least two different target relative positions, wherein, when the first road image and the second road image are at the target relative positions, at least one road element in the first road image and the second road image has an overlapping position;

[0143] The calculation submodule 8052 is also used to calculate the initial score value corresponding to the relative position of each target, so as to obtain at least two initial score values;

[0144] The calculation submodule 8052 is further configured to determine the initial score value with the highest score value among the at least two initial score values ​​as the target score value corresponding to the first modeling data.

[0145] Optionally, the acquisition module 801 includes:

[0146] The second generation submodule 8011 is used to generate a first grayscale image and a trajectory mask image based on an initial image, wherein the initial image is an image containing the road to be modeled, the first grayscale image is a grayscale image of the initial image, and the trajectory mask image includes trajectory information of the road to be modeled;

[0147] The recognition submodule 8012 is used to input the first grayscale image and the trajectory mask image into a pre-trained network model for semantic recognition, and obtain multiple semantic images output by the network model, wherein the semantic images include attribute parameters of one dimension of the base point;

[0148] The third generation submodule 8013 is used to generate the road data based on the multiple semantic images.

[0149] Optionally, the plurality of semantic images include directional semantic images, which include the initial direction of each pixel in the road to be modeled. The third generation submodule 8013 is used to correct the initial direction of the base point based on the initial direction of each pixel in the directional semantic image that is within a preset range from the base point, to obtain the target direction of the base point, wherein the road data includes the target direction.

[0150] Optionally, the road to be modeled includes multiple base points, and the multiple base points are multiple location points arranged at equal intervals along the extension direction of the road to be modeled. The location corresponding to the first base point is the road segment between the first base point and the second base point. The first base point is any base point among the multiple base points. The second base point is the base point among the multiple base points that is adjacent to the first base point.

[0151] It should be noted that the road model generation device 800 provided in this embodiment can realize all the technical solutions of the above-described road model generation method embodiment, and therefore can at least achieve all the above-described technical effects, which will not be repeated here.

[0152] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0153] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0154] Figure 13 A schematic block diagram of an example electronic device 1300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0155] like Figure 13As shown, the electronic device 1300 includes a computing unit 1301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1302 or a computer program loaded from a storage unit 1308 into a random access memory (RAM) 1303. The RAM 1303 may also store various programs and data required for the operation of the device 1300. The computing unit 1301, ROM 1302, and RAM 1303 are interconnected via a bus 1304. An input / output (I / O) interface 1305 is also connected to the bus 1304.

[0156] Multiple components in electronic device 1300 are connected to I / O interface 1305, including: input unit 1306, such as keyboard, mouse, etc.; output unit 1307, such as various types of displays, speakers, etc.; storage unit 1308, such as disk, optical disk, etc.; and communication unit 1309, such as network card, modem, wireless transceiver, etc. Communication unit 1309 allows device 1300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0157] The computing unit 1301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1301 performs the various methods and processes described above, such as the road model generation method. For example, in some embodiments, the road model generation method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1308. In some embodiments, part or all of the computer program may be loaded and / or installed on device 1300 via ROM 1302 and / or communication unit 1309. When the computer program is loaded into RAM 1303 and executed by the computing unit 1301, one or more steps of the road model generation method described above are performed. Alternatively, in other embodiments, the computing unit 1301 may be configured to perform a method for generating a road model by any other suitable means (e.g., by means of firmware).

[0158] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0159] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0160] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0161] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0162] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0163] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0164] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0165] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for generating a road model, comprising: Obtain road data corresponding to the base point in the road to be modeled, wherein the road data includes attribute parameters of the road element at the location corresponding to the base point; Based on the road data and the basic templates in the template library, multiple initial modeling data matching the road data are generated. The road elements corresponding to the basic templates match the road elements in the road to be modeled. The initial modeling data is used to generate the road model at the location corresponding to the base point. Target score values ​​are calculated for each of the initial modeling data based on the road data, and the target score values ​​are used to characterize the degree of matching between the road data and the base template; The base template corresponding to the initial modeling data with the highest target score among the multiple initial modeling data is determined as the target template corresponding to the road data; Based on the road data and the target template that matches the road data, target modeling data corresponding to the base point is generated. The target template includes road element models at the location corresponding to the base point. The target modeling data is used to generate a road model at the location corresponding to the base point. After determining the target template corresponding to the road data, the parameter scalar values ​​in the road to be modeled are identified and the parameter scalar values ​​are filled into the target template to obtain the target modeling data. A road model of the road to be modeled is generated based on the target modeling data, and the road model is used to generate map data.

2. The method according to claim 1, wherein, Based on the road data and basic templates in the template library, multiple initial modeling data matching the road data are generated, including: Identify at least two different base templates in the template library that match the road data; At least two initial modeling data that match the road data are generated based on each basic template, wherein each basic template corresponds to at least two initial modeling data. The multiple initial modeling data that match the road data include: at least two initial modeling data corresponding to the at least two different basic templates.

3. The method according to claim 2, wherein, The generation of at least two initial modeling data matching the road data based on each of the basic templates includes: The target parameters in the road data are scaled to obtain at least two first target road data, the at least two first target road data including: the road data and the road data after scaling the target parameters; For each of the at least two basic templates, at least two initial modeling data corresponding to the basic template are generated based on the basic template and the at least two first target road data.

4. The method according to claim 1, wherein, The plurality of initial modeling data includes first modeling data, which is initial modeling data generated based on the road data and a first basic template. The first basic template is a basic template in the template library. The step of calculating the target score value for each of the initial modeling data based on the road data includes: A first road image is determined based on the road data, and a second road image is determined based on the first base template; When the first road image and the second road image at least partially overlap, the target score value corresponding to the first modeling data is calculated based on the distance between each road element in the first road image and the corresponding road element in the second road image.

5. The method according to claim 4, wherein, When the first road image and the second road image at least partially overlap, the target score value corresponding to the first modeling data is calculated based on the distance between each road element in the first road image and the corresponding road element in the second road image, including: Determine at least two distinct target relative positions, wherein, when the first road image and the second road image are at the target relative positions, at least one road element in the first road image and the second road image has an overlapping position; Calculate the initial score value corresponding to the relative position of each target to obtain at least two initial score values; The initial score with the highest score among the at least two initial score values ​​is determined as the target score value corresponding to the first modeling data.

6. The method according to claim 1, wherein, The process of acquiring road data corresponding to base points in the road to be modeled includes: A first grayscale image and a trajectory mask image are generated based on an initial image, wherein the initial image is an image containing the road to be modeled, the first grayscale image is a grayscale image of the initial image, and the trajectory mask image includes the trajectory information of the road to be modeled; The first grayscale image and the trajectory mask image are input into a pre-trained network model for semantic recognition to obtain multiple semantic images output by the network model, wherein the semantic images include attribute parameters of one dimension of the base point; The road data is generated based on the multiple semantic images.

7. The method according to claim 6, wherein, The plurality of semantic images includes directional semantic images, wherein the directional semantic images include the initial direction of each pixel in the road to be modeled, and the generation of the road data based on the plurality of semantic images includes: Based on the initial directions of each pixel in the directional semantic image that is within a preset range from the base point, the initial direction of the base point is corrected to obtain the target direction of the base point, wherein the road data includes the target direction.

8. The method according to claim 1, wherein, The road to be modeled includes multiple base points, which are multiple location points arranged at equal intervals along the extension direction of the road to be modeled. The location corresponding to the first base point is the road segment between the first base point and the second base point. The first base point is any base point among the multiple base points. The second base point is the base point adjacent to the first base point among the multiple base points.

9. A road model generation apparatus, comprising: The acquisition module is used to acquire road data corresponding to the base point in the road to be modeled. The road data includes the attribute parameters of the road element at the location corresponding to the base point. The third generation module is used to generate multiple initial modeling data that match the road data based on the road data and the basic templates in the template library. The road elements corresponding to the basic templates match the road elements in the road to be modeled. The initial modeling data is used to generate the road model at the location corresponding to the base point. The calculation module is used to calculate a target score value for each of the initial modeling data based on the road data, wherein the target score value is used to characterize the degree of matching between the road data and the base template; The determining module is used to determine the base template corresponding to the initial modeling data with the highest target score among the multiple initial modeling data as the target template corresponding to the road data; The first generation module is used to generate target modeling data corresponding to the base point based on the road data and the target template matched with the road data. The target template includes road element models at the location corresponding to the base point. The target modeling data is used to generate a road model at the location corresponding to the base point. After determining the target template corresponding to the road data, the module identifies the parameter scalar values ​​in the road to be modeled and fills the parameter scalar values ​​into the target template to obtain the target modeling data. The second generation module is used to generate a road model of the road to be modeled based on the target modeling data, and the road model is used to generate map data.

10. The apparatus according to claim 9, wherein, The third generation module includes: The first determining submodule is used to determine at least two different basic templates in the template library that match the road data; The first generation submodule is used to generate at least two initial modeling data that match the road data based on each basic template, wherein each basic template corresponds to at least two initial modeling data. The multiple initial modeling data that match the road data include: at least two initial modeling data corresponding to the at least two different basic templates.

11. The apparatus according to claim 10, wherein, The first generation submodule is used to scale the target parameters in the road data to obtain at least two first target road data, wherein the at least two first target road data include: the road data and the road data after scaling the target parameters; The first generation submodule is further configured to generate at least two initial modeling data corresponding to each of the at least two basic templates, based on the basic template and the at least two first target road data.

12. The apparatus according to claim 9, wherein, The plurality of initial modeling data includes first modeling data, which is initial modeling data generated based on the road data and the first basic template. The first basic template is a basic template in the template library. The calculation module includes: The second determining submodule is used to determine a first road image based on the road data, and to determine a second road image based on the first base template; The calculation submodule is used to calculate the target score value corresponding to the first modeling data based on the distance between each road element in the first road image and the corresponding road element in the second road image when the first road image and the second road image at least partially overlap.

13. The apparatus according to claim 12, wherein, The calculation submodule is used to determine at least two different target relative positions, wherein, when the first road image and the second road image are at the target relative positions, at least one road element in the first road image and the second road image has an overlapping position; The calculation submodule is also used to calculate the initial score value corresponding to the relative position of each target, so as to obtain at least two initial score values; The calculation submodule is further configured to determine the initial score value with the highest score value among the at least two initial score values ​​as the target score value corresponding to the first modeling data.

14. The apparatus according to claim 9, wherein, The acquisition module includes: The second generation submodule is used to generate a first grayscale image and a trajectory mask image based on an initial image, wherein the initial image is an image containing the road to be modeled, the first grayscale image is a grayscale image of the initial image, and the trajectory mask image includes the trajectory information of the road to be modeled; The recognition submodule is used to input the first grayscale image and the trajectory mask image into a pre-trained network model for semantic recognition, and obtain multiple semantic images output by the network model, wherein the semantic images include attribute parameters of one dimension of the base point; The third generation submodule is used to generate the road data based on the multiple semantic images.

15. The apparatus according to claim 14, wherein, The plurality of semantic images include directional semantic images, which include the initial direction of each pixel in the road to be modeled. The third generation submodule is used to correct the initial direction of the base point based on the initial direction of each pixel in the directional semantic image that is within a preset range from the base point, so as to obtain the target direction of the base point. The road data includes the target direction.

16. The apparatus according to claim 9, wherein, The road to be modeled includes multiple base points, which are multiple location points arranged at equal intervals along the extension direction of the road to be modeled. The location corresponding to the first base point is the road segment between the first base point and the second base point. The first base point is any base point among the multiple base points. The second base point is the base point adjacent to the first base point among the multiple base points.

17. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed, enable the at least one processor to perform the steps of the road model generation method according to any one of claims 1-8.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the steps of the method for generating a road model according to any one of claims 1-8.

19. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method for generating a road model according to any one of claims 1-8.

Images