Construction method, device, system and equipment of pavement model and storage medium

Abstract

The invention relates to a pavement model construction method, device and system, equipment and a storage medium, and relates to the technical field of image processing. The method comprises the following steps: acquiring multiple pieces of road scene data of a target road; the road scene data comprises a plurality of image data acquired by the acquisition equipment and acquisition state information of the acquisition equipment; determining three-dimensional feature data based on the multiple pieces of road scene data; and constructing a pavement model of the target road based on the three-dimensional feature data of the road markers on the target road and the acquisition state information of the acquisition equipment. Therefore, the precision of three-dimensional reconstruction can be improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and more particularly to the field of three-dimensional image reconstruction, specifically to a method, apparatus, system, device, and storage medium for constructing a road surface model. Background Technology

[0002] In the current development of autonomous driving technology, the use of image data collected by vehicle-mounted cameras for 3D reconstruction of the road environment has become an indispensable key step. However, the diversity and dynamism of the road environment, including different weather conditions, lighting changes, road structure, and the behavior of traffic participants, all place high demands on the acquisition and processing of image data.

[0003] The related technology CN118365809A can determine the road reconstruction area and divide it into grid points. The target grid point information is then input into the reconstruction model to obtain 3D coordinates and semantic information. After verifying the accuracy of the vehicle position using vehicle pose, 3D information, and image optical flow, the model is used to perform semantic 3D reconstruction of the entire area. However, this technology fails to fully integrate multi-camera resources, which significantly limits the coverage and accuracy of the 3D reconstruction.

[0004] The related technology CN115937436A can utilize multiple roadside sensors to collect environmental data from different perspectives, preprocess the data to generate 3D model slices from each perspective, align them to the same coordinate system and correct distortion, and finally fuse them into a 3D model of the road scene. However, this method cannot achieve accurate 3D reconstruction of static models, resulting in poor reconstruction accuracy.

[0005] Therefore, the accuracy of road 3D reconstruction methods in related technologies is relatively poor. Summary of the Invention

[0006] This application provides a method, apparatus, system, device, and storage medium for constructing a road surface model, to at least solve the technical problem of poor accuracy in related three-dimensional road reconstruction methods. The technical solution of this application is as follows:

[0007] According to a first aspect provided in this application, a method for constructing a road surface model is provided, comprising: acquiring multiple road scene data of a target road; the road scene data including image data acquired by an acquisition device and acquisition status information of the acquisition device; determining feature matching data of the road scene data; and determining three-dimensional feature data of road signs on the target road based on the road scene data and the feature matching data; the feature matching data including at least one common-view matching pair; the image data included in the common-view matching pair having similar or identical image content; and constructing a road surface model of the target road based on the three-dimensional feature data and the acquisition status information.

[0008] Based on the aforementioned technical means, this application, through the acquisition of multiple image data by the data collection device, can capture various details of the road environment, providing a rich information source for subsequent data processing and analysis. Furthermore, by fusing multiple image data and pose information, a more realistic and detailed road surface model can be constructed, avoiding the problem of poor accuracy in 3D road reconstruction in related technologies. In addition, the constructed road surface model can be combined with autonomous driving technology to provide vehicles with accurate navigation and positioning information, improving driving comfort and safety.

[0009] One possible approach to determining feature matching data for road scene data includes: extracting features from image data in the road scene data to obtain pixel feature information of the image data; and determining feature matching data for the road scene data based on the pixel feature information.

[0010] Based on the aforementioned technical means, this application utilizes the pixel feature information of road signs in multiple road scene data to ensure that the matched image pairs have a high degree of similarity and overlap in content, so as to further determine the three-dimensional feature data of road signs on the target road, so as to accurately reflect the road features of the target road, thereby constructing a more realistic and detailed road surface model.

[0011] In one possible approach, based on road scene data and feature matching data, the three-dimensional feature data of road signs on the target road are determined, including: based on feature matching data and acquisition status information of multiple acquisition devices, three-dimensional feature reconstruction is performed on the road scene data to obtain multiple three-dimensional road data; the multiple three-dimensional road data are aligned to obtain three-dimensional feature data.

[0012] Based on the aforementioned technical means, this application can improve the accuracy of the overall data by integrating information from multiple data sources through alignment processing, thereby reducing the errors or deviations that may be caused by a single data source.

[0013] In one possible approach, three-dimensional feature reconstruction is performed on road scene data based on feature matching data and the acquisition status information of each of the multiple acquisition devices to obtain multiple three-dimensional road data. This includes: correcting the acquisition status information of each of the multiple acquisition devices based on their respective extrinsic parameters to obtain corrected acquisition status information of each of the multiple acquisition devices; and performing three-dimensional feature reconstruction on road scene data based on feature matching data and the corrected acquisition status information of each of the multiple acquisition devices to obtain multiple three-dimensional road data.

[0014] Based on the above technical means, this application can accurately correct the acquisition status information of each acquisition device by considering the external parameters of each acquisition device, eliminating data distortion or inconsistency caused by device position deviation, angle tilt or internal calibration error. Thus, when performing three-dimensional feature reconstruction based on the corrected acquisition status information and feature matching data of road scene data, it can more accurately construct the three-dimensional model of the road scene.

[0015] One possible approach involves aligning multiple 3D road data sets to obtain 3D feature data, including: determining shared field-of-view information for road scene data based on the corrected acquisition status information of each of the multiple acquisition devices; and aligning the multiple 3D road data sets based on the shared field-of-view information to obtain 3D feature data.

[0016] Based on the aforementioned technical means, this application can determine the shared field of view information among multiple road scene data by using the corrected acquisition status information of multiple acquisition devices, thereby achieving the alignment of multiple 3D road data. This eliminates the differences caused by data acquired by different devices or at different times, improving data consistency and comparability.

[0017] In one possible approach, a road surface model of the target road is constructed based on 3D feature data and the acquisition status information of the acquisition devices. This includes: updating the acquisition status information of multiple acquisition devices based on the 3D feature data to obtain updated acquisition status information for each of the multiple acquisition devices; performing semantic recognition on the 3D feature data based on the updated acquisition status information of each of the multiple acquisition devices to determine the pixel distribution of road signs on the target road in a 2D image space; and rendering the road signs on the target road based on the pixel distribution of the road signs on the target road in the 2D image space to obtain a road surface model of the target road.

[0018] Based on the above technical means, this application can utilize the three-dimensional feature data of road signs and combine it with the acquisition status information of the acquisition devices (such as LiDAR, cameras, etc.) to realize the pixel distribution of the road sign position in the two-dimensional image space, thereby accurately rendering the road signs on the target road and obtaining the road surface model of the target road.

[0019] In one possible approach, feature matching data of dynamic targets on the target road is determined based on road scene data; the feature matching data of dynamic targets is then rendered into the road surface model to obtain a three-dimensional dynamic model corresponding to the dynamic targets.

[0020] Based on the aforementioned technical means, this application can accurately capture the position, speed, direction and other feature data of dynamic targets (such as vehicles, pedestrians, non-motorized vehicles, etc.) on the road by processing multiple roads, so as to provide a powerful data support and visualization platform for autonomous driving technology research and development, intelligent transportation planning and design and other fields through three-dimensional dynamic models.

[0021] In one possible approach, the acquisition status information includes the pose of the acquisition device when acquiring image data, as well as the intrinsic and extrinsic parameter data of the acquisition device.

[0022] According to a second aspect provided in this application, a road surface model construction apparatus is provided, comprising: an acquisition unit, a determination unit, and a construction unit; the acquisition unit is used to acquire multiple road scene data of a target road; the road scene data includes image data acquired by an acquisition device and acquisition status information of the acquisition device; the determination unit is used to determine feature matching data of the road scene data, and based on the road scene data and the feature matching data, determine three-dimensional feature data of road signs on the target road; the feature matching data includes at least one common-view matching pair; the image data included in the common-view matching pair have similar or identical image content; the construction unit is used to construct a road surface model of the target road based on the three-dimensional feature data and the acquisition status information.

[0023] In one possible approach, the determining unit is specifically used for: extracting features from image data in road scene data to obtain pixel feature information of the image data; and determining feature matching data of the road scene data based on the pixel feature information.

[0024] In one possible approach, the unit is specifically used for: performing 3D feature reconstruction on road scene data based on feature matching data and acquisition status information from multiple acquisition devices to obtain multiple 3D road data; and aligning the multiple 3D road data to obtain 3D feature data.

[0025] In one possible approach, the unit is specifically used to: correct the acquisition status information of each of the multiple acquisition devices based on their respective external parameters, to obtain corrected acquisition status information of each of the multiple acquisition devices; and perform three-dimensional feature reconstruction on the road scene data based on feature matching data and the corrected acquisition status information of each of the multiple acquisition devices, to obtain multiple three-dimensional road data.

[0026] In one possible approach, the determining unit is specifically used for: aligning multiple 3D road data to obtain 3D feature data, including: determining shared field-of-view information of road scene data based on the corrected acquisition status information of each of the multiple acquisition devices; and aligning multiple 3D road data based on the shared field-of-view information to obtain 3D feature data.

[0027] In one possible approach, the construction unit is specifically used for: updating the acquisition status information of multiple acquisition devices based on 3D feature data to obtain updated acquisition status information of each acquisition device; performing semantic recognition on the 3D feature data based on the updated acquisition status information of each acquisition device to determine the pixel distribution of road signs on the target road in the 2D image space; and rendering the road signs on the target road based on the pixel distribution of road signs on the target road in the 2D image space to obtain a road surface model of the target road.

[0028] In one possible approach, the determining unit is also used to determine the feature data of dynamic targets on the target road based on road scene data; the constructing unit is used to render the feature data of the dynamic targets into the road surface model to obtain the three-dimensional dynamic model corresponding to the dynamic targets.

[0029] According to a third aspect provided in this application, a road surface model construction system is provided. The system includes: a computing resource scheduler, a data resource manager, and multiple processors. The computing resource scheduler is used to acquire multiple road scene data of a target road from the data resource manager. The road scene data includes multiple image data acquired by an acquisition device, as well as acquisition status information of the acquisition device. The scheduler is also used to determine a target processor based on the resource usage of each processor and send multiple road scene data to the target processor. The target processor is the processor among the multiple processors whose resource occupancy rate is less than a preset threshold. The target processor is used to: determine three-dimensional feature data based on the road scene data and feature matching data; the feature matching data includes multiple common-view matching pairs; the common-view matching pairs include multiple image data with similar or identical image content; and construct a road surface model of the target road based on the three-dimensional feature data and the acquisition status information of the acquisition device.

[0030] According to a fourth aspect provided in this application, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute instructions to implement the method of the first aspect described above and any possible implementation thereof.

[0031] According to a fifth aspect provided in this application, a computer-readable storage medium is provided that, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the methods described in the first aspect and any possible implementation thereof.

[0032] According to the sixth aspect provided in this application, a computer program product is provided, the computer program product including computer instructions, which, when executed on an electronic device, cause the electronic device to perform the method described in the first aspect and any possible implementation thereof.

[0033] Therefore, the above-mentioned technical features of this application have the following beneficial effects:

[0034] (1) Multiple image data acquired by the acquisition device can capture various details of the road environment, providing a rich source of information for subsequent data processing and analysis. In addition, by fusing multiple image data and pose information, a more realistic and detailed road surface model can be constructed, avoiding the problem of poor accuracy in road 3D reconstruction in related technologies. Furthermore, the constructed road surface model can be combined with autonomous driving technology to provide vehicles with accurate navigation and positioning information, improving driving comfort and safety.

[0035] (2) By utilizing the pixel feature information of road signs in multiple road scene data, the matching image pairs are ensured to have a high degree of similarity and overlap in content, so as to further determine the three-dimensional feature data and accurately reflect the road features of the target road, thereby constructing a more realistic and detailed road surface model.

[0036] (3) Alignment processing can integrate information from multiple data sources, reduce errors or deviations that may be caused by a single data source, and thus improve the accuracy of the overall data.

[0037] (4) By considering the external parameters of each acquisition device, the acquisition status information of each device can be accurately corrected, eliminating data distortion or inconsistency caused by device position deviation, angle tilt or internal calibration error. Thus, based on the corrected acquisition status information and feature matching data of road scene data, when performing three-dimensional feature reconstruction, the three-dimensional model of the road scene can be constructed more accurately.

[0038] (5) By using the corrected acquisition status information of multiple acquisition devices, the shared field of view information among multiple road scene data can be determined, thereby achieving the alignment of multiple 3D road data. This eliminates the differences caused by data acquisition from different devices or at different times, and improves the consistency and comparability of the data.

[0039] (6) By utilizing the three-dimensional feature data of road signs and combining the acquisition status information of acquisition devices (such as lidar, cameras, etc.), the pixel distribution of the location of road signs in the two-dimensional image space can be realized, thereby accurately rendering the road signs on the target road and obtaining the road surface model of the target road.

[0040] (7) By processing multiple roads, the location, speed, direction and other feature data of dynamic targets on the road (such as vehicles, pedestrians, non-motorized vehicles, etc.) can be accurately captured, so as to provide a powerful data support and visualization platform for autonomous driving technology research and development, intelligent transportation planning and design through three-dimensional dynamic models.

[0041] It should be noted that the technical effects of any of the implementation methods in aspects two through five can be found in the technical effects of the corresponding implementation methods in aspect one, and will not be repeated here.

[0042] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0043] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application, and do not constitute an undue limitation of this application.

[0044] Figure 1 This is a flowchart illustrating a method for constructing a road surface model according to an exemplary embodiment;

[0045] Figure 2 This is a schematic diagram illustrating the structure of a road surface model construction system according to an exemplary embodiment;

[0046] Figure 3 This is a schematic diagram illustrating a single road scene data reconstruction process according to an exemplary embodiment;

[0047] Figure 4 This is a schematic diagram illustrating a process for aligning multiple road scene data according to an exemplary embodiment;

[0048] Figure 5 This is a schematic diagram illustrating a static model construction process according to an exemplary embodiment;

[0049] Figure 6 This is a schematic diagram illustrating a dynamic model construction process according to an exemplary embodiment;

[0050] Figure 7 This is a schematic diagram illustrating a computing resource scheduler according to an exemplary embodiment;

[0051] Figure 8 This is a schematic diagram illustrating the effect of a road surface model according to an exemplary embodiment;

[0052] Figure 9 This is a schematic diagram illustrating the effect of a dynamic target according to an exemplary embodiment;

[0053] Figure 10 This is a schematic diagram illustrating the effect of a three-dimensional dynamic model according to an exemplary embodiment;

[0054] Figure 11This is a block diagram illustrating a road surface model construction apparatus according to an exemplary embodiment;

[0055] Figure 12 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0056] To enable those skilled in the art to better understand the technical solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0057] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0058] As mentioned in the background section, to address the issues of poor accuracy and low efficiency in related road 3D reconstruction methods, this application provides a method for constructing a road surface model. This method acquires multiple road scene data of the target road and determines 3D feature data based on this data. Furthermore, it constructs a road surface model of the target road based on the 3D feature data of road signs on the target road and the pose of the image data acquired by the acquisition device. This allows for efficient and accurate 3D reconstruction of the target road.

[0059] To facilitate understanding, the method for constructing the road surface model provided in this application will be described in detail below with reference to the accompanying drawings.

[0060] Figure 1 This is a flowchart illustrating a method for constructing a road surface model according to an exemplary embodiment, such as... Figure 1 As shown, the method for constructing this road surface model includes the following steps: S101-S103.

[0061] S101, The road surface model construction device acquires multiple road scene data of the target road.

[0062] The road scene data may include image data collected by the acquisition device, as well as the acquisition status information of the acquisition device.

[0063] In one possible approach, the image data can be image data collected by one or more acquisition devices while multiple mobile vehicles are traveling on the target road, or it can be image data collected by one or more acquisition devices while a single mobile vehicle is traveling on the target road. This application does not impose any specific limitations on this.

[0064] Optionally, the mobile carrier can be set according to actual needs. For example, the mobile carrier can be a vehicle or an aircraft. This application does not impose specific limitations in this regard.

[0065] In one example, the multiple road scene data for the target road can be collected by multiple acquisition devices, or by a single acquisition device. This application does not impose specific limitations in this regard.

[0066] In one possible approach, the road surface model building device can be configured with a storage module. The storage module can store road scene data for multiple roads. The road surface model building device can read multiple road scene data for the target road from the storage module. The storage module can be configured with a data filter. Based on the data filter, the storage module can filter the road scene data for multiple roads to obtain multiple road scene data for the target road.

[0067] Optionally, the data acquisition device can be configured according to actual needs. For example, the data acquisition device can be a monocular camera or a radar. This application does not impose specific limitations in this regard.

[0068] In one possible approach, the acquisition status information includes the pose of the acquisition device when acquiring image data, as well as the intrinsic parameters of the acquisition device. The pose of the acquisition device when acquiring image data includes its position and orientation information. The intrinsic parameter data of the acquisition device includes its focal length, camera optical coordinates, distortion parameters, etc. The extrinsic parameter data of the acquisition device includes translation and rotation parameters between acquisition devices, as well as translation and rotation parameters between the acquisition device and different coordinate systems, etc.

[0069] In one example, road scene data may include multiple image data acquired by a monocular camera, as well as inertial measurement unit (IMU) data, global navigation satellite system (GNSS) data, and monocular camera calibration data. The road surface model construction device can determine the pose of the monocular camera when acquiring image data based on the IMU data, GNSS data, and monocular camera calibration data. The monocular camera calibration data can be used to characterize the calibrated pose of the monocular camera on the vehicle.

[0070] S102, The road surface model construction device determines the feature matching data of the road scene data, and determines the three-dimensional feature data of the road signs on the target road based on the road scene data and the feature matching data.

[0071] The feature matching data may include multiple co-view matching pairs. A co-view matching pair may include multiple image data that have similar or identical image content;

[0072] In one possible approach, the road surface model construction device can determine feature matching data from the road scene data. The device can extract features from the image data within the road scene data to obtain pixel feature information. Based on this pixel feature information, the device can determine the feature matching data for the road scene data.

[0073] Specifically, the road surface model construction device can employ advanced image recognition and processing technologies to extract features from multiple image datasets, automatically identify and label traffic elements such as road boundaries, road markings, road surface cracks, potholes, and water accumulation in the images, and generate corresponding pixel feature information. These image recognition and processing technologies include, but are not limited to, convolutional neural networks and feature pyramid networks.

[0074] The road surface model construction device can identify and pair images that are similar or overlapping in content but were taken from different viewpoints, locations, or times after obtaining the pixel features of multiple image data. These successfully matched image pairs have obvious overlap in their pixel features.

[0075] Because different acquisition devices may have positional deviations, angular tilts, or time differences, directly using common-view matching pairs for model construction may lead to error accumulation. Therefore, the road surface model construction device can correct multiple common-view matching pairs based on the acquisition status information of multiple acquisition devices to obtain accurate feature matching data for road scene data.

[0076] In one possible approach, the road surface model construction device can reconstruct three-dimensional features from road scene data based on feature matching data and acquisition status information from multiple acquisition devices, obtaining multiple three-dimensional road data. The road surface model construction device can then align these multiple three-dimensional road data to obtain three-dimensional feature data. The specific implementation method by which the road surface model construction device determines the three-dimensional feature data of road signs on the target road based on the road scene data and feature matching data can be found in S201-S202 below. Further details are omitted here.

[0077] S103. The road surface model construction device constructs a road surface model of the target road based on three-dimensional feature data and collected status information.

[0078] In one possible approach, the road surface model construction device can update the acquisition status information of multiple acquisition devices based on 3D feature data, obtaining updated acquisition status information for each device. The device can then perform semantic recognition on the 3D feature data based on the updated acquisition status information of the multiple acquisition devices to determine the pixel distribution of road signs on the target road in a 2D image space. Finally, based on the pixel distribution of road signs on the target road in the 2D image space, the device can render the road signs on the target road to obtain a road surface model of the target road.

[0079] Specifically, the road surface model construction device integrates semantic recognition technology and image rendering technology. Based on semantic recognition technology, the device can accurately identify the three-dimensional feature data of various road markings in image data and calculate the pixel distribution of the road markings in the two-dimensional image space.

[0080] The road surface model building device can use graphics rendering technology to render the pixel distribution of road markers in a two-dimensional image space to obtain the road surface rendering result of the target road. The device can then integrate the road surface rendering results of the target road to generate a highly realistic road surface model containing rich details.

[0081] In one possible approach, road surface models can be used in multiple fields such as urban planning, traffic simulation, and autonomous driving testing. They can also provide crucial data support and decision-making basis for road design, maintenance, and safety management. The road surface model building device can also update the model data in real time based on the acquired data to reflect actual changes and dynamic information about the road.

[0082] In another possible approach, the road surface model building device can determine the feature data of dynamic targets on the target road based on road scene data. The device can then render the feature data of the dynamic targets into the road surface model to obtain a three-dimensional dynamic model corresponding to the dynamic targets.

[0083] Specifically, the road surface model building device can identify dynamic targets on a target road, such as pedestrians and vehicles, based on multiple road scene data. The device can analyze key information such as the position, speed, direction, size, and even behavior patterns of dynamic targets on the target road based on multiple road scene data, thereby extracting the feature data of the dynamic targets.

[0084] After obtaining the feature data of a dynamic target, the road surface model building device can integrate this data into a pre-constructed road surface model using a graphics rendering engine and 3D modeling technology. Through refined texture mapping, lighting and shadow effects processing, and dynamic physical simulation, the road surface model building device can generate and display the corresponding 3D dynamic model of the dynamic target in real time.

[0085] The three-dimensional dynamic model constructed by the road surface model building device can provide an intuitive and comprehensive tool for monitoring and analyzing traffic conditions, and also provides strong data support and a visualization platform for fields such as autonomous driving technology research and development, intelligent transportation planning and design.

[0086] based on Figure 1 The technical solution described in this application utilizes multiple image data acquired by a data acquisition device to capture various details of the road environment, providing a rich information source for subsequent data processing and analysis. Furthermore, by fusing multiple image data and pose information, a more realistic and detailed road surface model can be constructed, avoiding the problem of poor accuracy in 3D road reconstruction in related technologies. In addition, the constructed road surface model can be combined with autonomous driving technology to provide vehicles with accurate navigation and positioning information, improving driving comfort and safety.

[0087] In some embodiments, in order to determine the three-dimensional feature data of road signs on a target road based on multiple road scene data and feature matching data of multiple road scene data, the road surface model construction apparatus provided in this application embodiment further includes the following steps: S201-S202.

[0088] S201, the road surface model construction device reconstructs three-dimensional features from road scene data based on feature matching data and the acquisition status information of multiple acquisition devices, and obtains multiple three-dimensional road data.

[0089] In one possible approach, the road surface model construction device can correct the acquisition status information of each of the multiple acquisition devices based on their respective extrinsic parameters, thereby obtaining corrected acquisition status information for each of the multiple acquisition devices. The road surface model construction device can then perform 3D feature reconstruction of the road scene data based on feature matching data and the corrected acquisition status information of each of the multiple acquisition devices, resulting in multiple 3D road data sets.

[0090] Specifically, the staff can store the extrinsic parameters of multiple acquisition devices in the storage module configured in the road surface model building device. The road surface model building device can read the extrinsic parameters of each acquisition device from the storage module. Based on the extrinsic parameters of each acquisition device, the road surface model building device can correct the acquisition status information of each acquisition device, obtaining corrected acquisition status information for each acquisition device. The corrected acquisition status information is more accurate.

[0091] In one possible approach, the road surface model construction device can perform three-dimensional feature reconstruction on the image data in the road scene data based on feature matching data of the road scene data and the corrected acquisition state information of multiple acquisition devices, that is, to correct the image data to obtain the three-dimensional road data corresponding to the road scene data.

[0092] In one example, the road surface model construction device can use sophisticated image processing and data analysis algorithms to identify three-dimensional features of road signs, such as their location, shape, color, and size, as well as road surface features such as geometric shape, slope, and curvature, in road scene data. Based on the corrected acquisition status information from multiple acquisition devices, the device can reconstruct three-dimensional features from these road scene data to obtain accurate three-dimensional road data.

[0093] In another possible approach, the road surface model building device can be based on a composite matching filtering algorithm to filter out erroneous 3D road data.

[0094] S202, the road surface model construction device aligns multiple three-dimensional road data to obtain three-dimensional feature data of road signs on the target road.

[0095] In one possible approach, the road surface model construction device can determine shared field-of-view information for road scene data based on the corrected acquisition status information of multiple acquisition devices. Based on this shared field-of-view information, the device can then align multiple 3D road data sets to obtain 3D feature data.

[0096] Specifically, the road surface model construction device can use the corrected acquisition status information of multiple acquisition devices to determine the shared field of view information among multiple road scene data, that is, the similar or identical parts of the image data in the multiple road scene data. Based on the shared field of view information of multiple road scene data, the road surface model construction device can align and fuse multiple 3D road data, and merge the multiple 3D road data together to obtain the 3D feature data of road signs on the target road.

[0097] Based on this, this application can improve the accuracy of the overall data by integrating information from multiple data sources through alignment processing, reducing the errors or deviations that may be caused by a single data source.

[0098] In some embodiments, such as Figure 2 The diagram shown is a schematic diagram of a road surface model construction system provided in this application.

[0099] In one possible approach, the road surface model building system includes a data filter, a road surface model building device and a scene simulation simulator, a data resource manager, and a computing resource scheduler.

[0100] The data filtering device can filter multiple road scene data to obtain multiple road scene data corresponding to the target road. For example, road scene data 1, road scene data 2, road scene data 3, ..., road scene data N.

[0101] The road surface model construction device may include a data acquisition module, a single road scene reconstruction module, a multi-road scene alignment module, a static model construction module, and a dynamic model construction module.

[0102] The data acquisition module can obtain multiple initial road scene data on the target road from the data filtering device. The data acquisition module can process the initial road scene data recorded by each vehicle on the target road according to the configured data extraction component and data merging component, thereby obtaining multiple road scene data on the target road.

[0103] The single road scene reconstruction module can be configured with feature extraction components, field-of-view sharing and matching components, extrinsic parameter calibration components, pose initialization components, and reconstruction and optimization components. Based on these configured components, the module processes each road scene data from multiple road scene datasets to obtain the reconstruction results for each data point. These results include the feature data corresponding to each road scene data point and the accurate pose of the acquisition device.

[0104] The multi-road scene alignment module can be configured with a field-of-view sharing component and a trajectory alignment component. Based on the configured field-of-view sharing component and trajectory alignment component, the multi-road scene alignment module can match and align the feature data corresponding to each road scene data to obtain the optimized reconstruction results of all road scene data, namely the 3D feature data of road signs on the target road, and the accurate acquisition status information of each acquisition device in each optimized driving data.

[0105] The static model building module can be configured with feature fusion components and scene inference components. Based on the configured feature fusion components and scene inference components, the static model building module can process the alignment results of multiple road scenes and the accurate pose of the acquisition devices to build a road surface model of the target road.

[0106] The dynamic model building module can be configured with a dynamic target recovery component and a scene fusion component. Based on the configured dynamic target recovery component and scene fusion component, the dynamic model building module can process the static road surface model of the target road and multiple road scene data on the target road to construct a 3D dynamic model.

[0107] The data resource manager is configured with road scene data storage, reconstruction data storage, alignment data storage, static model storage, and dynamic model storage.

[0108] The system includes: a road scene data storage unit for storing road scene data; a reconstruction data storage unit for storing reconstruction data for a single road scene; an alignment data storage unit for storing alignment data for multiple road scenes; a static model storage unit for storing static models; and a dynamic model storage unit for storing dynamic models.

[0109] The computational resource scheduler can be used to retrieve multiple road scene data for a target road from the data resource manager. The scheduler can also determine a target processor based on the resource usage of each processor and send multiple road scene data to that target processor; the target processor is the processor among the multiple processors whose resource utilization is less than a preset threshold. The target processor can be used for:

[0110] Based on multiple road scene data and feature matching data of multiple road scene data, the three-dimensional feature data of road signs on the target road are determined; the feature matching data includes multiple co-view matching pairs; the co-view matching pairs include multiple image data with similar or identical image content;

[0111] Based on the three-dimensional feature data of road markers on the target road and the acquisition status information of the acquisition equipment, a road surface model of the target road is constructed.

[0112] Optionally, the preset threshold can be set according to actual needs. For example, the preset threshold can be 40% or 20%. This application does not impose specific limitations in this regard. Figure 2 ,like Figure 3 The diagram shown is a schematic of a single road scene reconstruction process.

[0113] In one possible approach, a single road scene reconstruction module can obtain road scene data from a data resource manager, including multiple image data and acquisition status information of the acquisition device.

[0114] The feature extraction component in the single road scene reconstruction module can employ feature extraction algorithms to extract features from the road scene data, obtaining the pixel features of each image data point in the road scene data. The feature extraction component uses a composite feature extraction algorithm to extract features from the image, which includes a deep learning algorithm and a scale-invariant feature transform (SIFT) algorithm. For ordinary images, an improved deep learning algorithm is used to improve processing efficiency, and for scenes with large lighting variations, the SIFT algorithm ensures effective extraction of image features.

[0115] In the single road scene reconstruction module, the field-of-view sharing and matching component can utilize a field-of-view sharing algorithm to identify and integrate shared field-of-view data from different cameras on the same vehicle at the same or different times. Combined with feature extraction results, an optimized deep learning feature matching algorithm is used to achieve efficient data matching and enhance the coherence between data points.

[0116] Given the complexity of the driving environment, which may cause fluctuations in camera extrinsic parameters, the extrinsic parameter calibration component can monitor and identify extrinsic parameter deviations through a sophisticated extrinsic parameter verification algorithm, and then apply an extrinsic parameter self-calibration algorithm for real-time correction to ensure the accuracy of camera parameters.

[0117] The pose initialization component can quickly and accurately reconstruct the 3D feature data of a single vehicle's driving data based on an improved Structure from Motion (SFM) algorithm and extrinsic constraint algorithm. This process not only considers the rigid connections between cameras, but also further optimizes the reconstruction results through extrinsic constraint algorithms, improving the accuracy of the camera's acquisition state information.

[0118] After reconstruction is completed, multi-camera reconstruction algorithms and a distributed optimization strategy are used to deeply fuse and optimize the reconstruction results of each camera, ensuring that the final reconstruction result of a single road scene is both accurate and complete.

[0119] In one possible approach, the design of a single road scene reconstruction module offers high flexibility and scalability. The components are independent of each other and can be flexibly combined according to actual needs. For example, they can be processed according to a pre-defined workflow, or the processing order can be adjusted based on the specific scenario.

[0120] To optimize data processing efficiency and reduce hardware burden in the single road scene reconstruction module, a data flow management mechanism was designed to output the processing results of each component to the data resource manager in real time and store them in the single road scene reconstruction data storage, effectively reducing real-time computing pressure and improving the overall system performance.

[0121] Combination Figure 2 ,like Figure 4 The diagram shown is a schematic of a process for aligning data from multiple road scenes.

[0122] In one possible approach, the multi-road scene alignment module can obtain single-road scene reconstruction data from the data resource manager, namely, feature matching data of the road scene data and acquisition status information of the acquisition device.

[0123] The field-of-view sharing component configured in the multi-road scene alignment module can acquire shared field-of-view information from image data acquired by different cameras at different time points, spatial locations, and from multiple road scene data points using a field-of-view sharing algorithm, thus obtaining shared field-of-view data for the target road. Furthermore, a composite matching filtering algorithm efficiently removes erroneous shared field-of-view data, ensuring the accuracy of subsequent processing.

[0124] Based on accurate shared field-of-view data, the trajectory alignment component can employ a fast error estimation algorithm to quickly analyze and quantify the deviation between 3D spatial points and corresponding image feature points, thereby selecting a set of feature points with low error. Furthermore, a high-precision alignment algorithm is used to achieve fine alignment of the reconstructed results of each trajectory data, ensuring high quality and consistency of global data fusion.

[0125] In one possible approach, the multi-road scene alignment module is designed with high flexibility and scalability. The components are independent of each other and can be flexibly combined according to actual needs. For example, it can process according to a set process sequence, or the processing sequence can be adjusted according to the specific scenario.

[0126] To improve the data processing efficiency of the multi-road scene alignment module and reduce the hardware burden, a data flow management mechanism was designed to output the processing results of each component to the data resource manager in real time and store them in the alignment data storage, which effectively reduced the real-time computing pressure and improved the overall system performance.

[0127] Combination Figure 2 ,like Figure 5 The diagram shown is a schematic of a static model construction process.

[0128] In one possible approach, the static model building device can obtain multi-road scene alignment data from a data resource manager, namely, the 3D feature data of road signs on the target road and the acquisition status information of the acquisition device.

[0129] The feature fusion component configured in the static model building module can accurately remove invalid or redundant road feature information according to its built-in feature filtering algorithm to ensure data quality. It also integrates the three-dimensional feature data of road signs on the target road through a fast feature fusion algorithm to form a complete and accurate feature description of the global scene.

[0130] After feature fusion, the scene inference component configured in the static model building module can use advanced scene recognition algorithms to accurately identify special road features in the global scene, such as traffic signs and lane lines, and set map feature reconstruction targets accordingly. Subsequently, based on the poses of each acquisition device, these features are re-rendered into corresponding images. On this basis, the scene reconstruction algorithm can perform deep feature inference and semantic merging on the two-dimensional road environment data, accurately reconstruct ground features, and finally aggregate them into road surface data.

[0131] In one possible approach, the static model building module is designed with high flexibility and scalability. The components are independent of each other and can be flexibly combined according to actual needs. For example, they can be processed according to a set process sequence, or the processing sequence can be adjusted according to the specific scenario.

[0132] To optimize the data processing efficiency of the static model building module and reduce the hardware burden, a data flow management mechanism was designed to output the processing results of each component to the data resource manager in real time and store them in the road surface model memory, which effectively reduced the real-time computing pressure and improved the overall system performance.

[0133] Combination Figure 2 ,like Figure 6 The diagram shown is a schematic of a dynamic model construction process.

[0134] In one possible approach, the dynamic model building device can obtain road scene data and road surface models from a data resource manager.

[0135] The dynamic target recovery component configured in the dynamic reconstruction module can accurately locate and identify dynamic targets at specified positions using dynamic target recognition algorithms. Subsequently, through an efficient tracking mechanism, it continuously captures and records the positional information of these dynamic targets at different image frames and time points. This allows for further in-depth analysis and reconstruction of the motion trajectory of the dynamic targets based on motion reconstruction algorithms, generating accurate representations of the dynamic target's motion state and motion model.

[0136] The dynamic reconstruction module can embed dynamic targets into corresponding static scenes based on motion recovery algorithms. Similarly, the dynamic reconstruction module can use scene rendering algorithms to perform fine-grained rendering processing on dynamic and static scene data, which is used for rendering the original model data and generating data for simulating new scenes.

[0137] The dynamic model building module can combine the model data obtained after the dynamic target reconstruction is completed with the road surface model. Through further processing by the scene fusion component, the coordinate system and scale standard are unified, and a scene representation with a clear structure and rigorous logic is constructed.

[0138] In one possible approach, the dynamic model building module is designed with high flexibility and scalability, with each component operating independently and capable of being flexibly combined according to actual needs. For example, it can be processed according to a pre-defined workflow sequence, or the processing order can be adjusted based on specific scenarios.

[0139] To optimize the data processing efficiency of the dynamic model building module and reduce the hardware burden, a data flow management mechanism was designed to output the processing results of each component to the data resource manager in real time and store them in the dynamic model memory, which effectively reduced the real-time computing pressure and improved the overall system performance.

[0140] Combination Figure 2 ,like Figure 7 The diagram shown is a schematic of a computing resource scheduler provided in this application.

[0141] In one possible approach, the computing resource scheduler may include: a cluster resource service architecture, a system resource service architecture, and a global topology management architecture. These three architectures can work collaboratively through process scheduling mechanisms and network connections to jointly drive the execution of the road model construction system.

[0142] The cluster resource service architecture serves as the foundational support for the system, comprising cluster nodes and a processing progress monitoring and management service. Cluster nodes can support distributed reconstruction across multiple devices, ensuring the successful completion of complex reconstruction tasks through efficient collaboration. The processing progress monitoring and management service not only monitors data progress in real time during data processing but also handles issue logging and resource monitoring.

[0143] The system resource service architecture includes: a runtime example management service, a computing resource monitoring and management service, and a data resource manager. The runtime example management service comprehensively monitors the running status of all work nodes, data flow transmission, and process construction details, ensuring the orderly operation of the entire system. The computing resource monitoring and management service assesses the status of computing resources and rationally allocates system resources according to the operational needs of nodes. The data resource manager includes a road scene data storage module and a reconstruction data storage module. The road scene data storage module stores and manages road scene data. The reconstruction data storage module stores the data after reconstruction.

[0144] A global topology management architecture is key to achieving efficient data processing and reconstruction. This architecture includes single-road scene worker nodes, multi-road scene worker nodes, and dynamic / static scene worker nodes. These nodes, scheduled by the running instance management service, collectively complete in-depth data processing and reconstruction tasks. Processing results are stored in the data resource manager. Single-road scene worker nodes process data for a single road scene. The processing flow includes data acquisition and data reconstruction. Multi-road scene worker nodes process data for multiple road scenes. The processing flow includes multi-road scene data alignment and static reconstruction. Dynamic / static scene worker nodes process road surface models and data from multiple road scenes. The processing flow includes dynamic reconstruction and scene fusion.

[0145] In one example, combining Figure 5 ,like Figure 8 The image shown is a schematic diagram of a road surface model. Figure 8 It includes the rendering results of the target road based on multiple road scene data, as well as the road surface model of the target road.

[0146] In one example, combining Figure 6 ,like Figure 9 The image shown is a schematic diagram illustrating the effect of a dynamic target. Figure 9 The image shows the position of the dynamic target at time 1, time 2, and time 3.

[0147] In one example, combining Figure 6 As shown in Figure 10, this is a schematic diagram of the effect of a three-dimensional dynamic model. The road surface model construction device can simulate dynamic targets in the real world into a three-dimensional dynamic model.

[0148] The above primarily describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the road surface model construction device or electronic device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0149] This application embodiment can, based on the above method, exemplarily divide the road surface model building device or electronic device into functional modules. For example, the road surface model building device or electronic device may include functional modules corresponding to each functional division, or two or more functions may be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this application embodiment is illustrative and only represents one logical functional division; in actual implementation, there may be other division methods.

[0150] Figure 11 This is a block diagram illustrating a road surface model construction apparatus according to an exemplary embodiment. (Refer to...) Figure 11 The road surface model construction device includes: an acquisition unit 401, a determination unit 402, and a construction unit 403.

[0151] In one possible approach, the acquisition unit 401 is used to acquire multiple road scene data of the target road; the road scene data includes image data acquired by the acquisition device, as well as the acquisition status information of the acquisition device.

[0152] In one possible approach, the determining unit 402 is used to determine the feature matching data of the road scene data, and based on the road scene data and the feature matching data, to determine the three-dimensional feature data of road signs on the target road.

[0153] In one possible approach, the building unit 403 is used to build a road surface model of the target road based on three-dimensional feature data and acquired state information.

[0154] In one possible approach, the determining unit 402 is specifically used to: extract features from image data in road scene data to obtain pixel feature information of the image data; and determine feature matching data of road scene data based on the pixel feature information.

[0155] In one possible approach, the determining unit 402 is specifically used to: perform three-dimensional feature reconstruction on road scene data based on feature matching data and acquisition status information of multiple acquisition devices to obtain multiple three-dimensional road data; and perform alignment processing on the multiple three-dimensional road data to obtain three-dimensional feature data.

[0156] In one possible approach, the determining unit 402 is specifically used to: correct the acquisition status information of each of the multiple acquisition devices based on their respective external parameters, to obtain corrected acquisition status information of each of the multiple acquisition devices; and perform three-dimensional feature reconstruction on the road scene data based on feature matching data and the corrected acquisition status information of each of the multiple acquisition devices, to obtain multiple three-dimensional road data.

[0157] In one possible approach, the determining unit 402 is specifically used to: perform alignment processing on multiple three-dimensional road data to obtain three-dimensional feature data, including: determining shared field-of-view information of road scene data based on the corrected acquisition status information of each of the multiple acquisition devices; and performing alignment processing on multiple three-dimensional road data based on the shared field-of-view information to obtain three-dimensional feature data.

[0158] In one possible approach, the construction unit 403 is specifically used for: updating the acquisition status information of each of the multiple acquisition devices based on the three-dimensional feature data to obtain the updated acquisition status information of each of the multiple acquisition devices; performing semantic recognition on the three-dimensional feature data based on the updated acquisition status information of each of the multiple acquisition devices to determine the pixel distribution of road signs on the target road in the two-dimensional image space; and rendering the road signs on the target road based on the pixel distribution of the road signs on the target road in the two-dimensional image space to obtain the road surface model of the target road.

[0159] In one possible approach, the determining unit 402 is also used to determine the feature data of dynamic targets on the target road based on road scene data.

[0160] In one possible approach, the building unit 403 is used to render the feature data of the dynamic target into the road surface model to obtain the three-dimensional dynamic model corresponding to the dynamic target.

[0161] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0162] Figure 12 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Figure 12 As shown, the electronic device includes, but is not limited to, a processor 501 and a memory 502.

[0163] The memory 502 described above is used to store the executable instructions of the processor 501. It is understood that the processor 501 is configured to execute instructions to implement the road surface model construction method in the above embodiments.

[0164] It should be noted that those skilled in the art will understand that Figure 12 The electronic device structure shown does not constitute a limitation on the electronic device; the electronic device may include, but is not limited to, other electronic devices. Figure 12 This may indicate more or fewer components, or combinations of certain components, or different component arrangements.

[0165] Processor 501 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in memory 502, and by calling data stored in memory 502, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Processor 501 may include one or more processing units. Optionally, processor 501 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into processor 501.

[0166] The memory 502 can be used to store software programs and various data. The memory 502 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required by at least one functional module (such as a determination unit, processing unit, etc.), etc. Furthermore, the memory 502 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0167] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 502 including instructions, which can be executed by a processor 501 of an electronic device to implement the methods in the above embodiments.

[0168] In actual implementation, Figure 11 The functions of the acquisition unit 401, the determination unit 402, and the processing unit 403 can all be provided by... Figure 12 The processor 501 calls the computer program stored in the memory 502 to implement the process. The specific execution process can be found in the description of the method section in the previous embodiment, and will not be repeated here.

[0169] Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0170] In an exemplary embodiment, this application also provides a computer program product including one or more instructions, which can be executed by a processor 501 of an electronic device to perform the methods described above.

[0171] It should be noted that when one or more instructions in the computer-readable storage medium or computer program product are executed by the processor of an electronic device, they implement the various processes of the above method embodiments and achieve the same technical effect as the above method. To avoid repetition, they will not be described again here.

[0172] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0173] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0174] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the classified units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0175] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0176] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, essentially, or the part that contributes to the prior art, or a complete or partial classification of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0177] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method of constructing a road surface model, characterized by, The method comprises: acquiring a plurality of road scene data of a target road; the road scene data comprises image data collected by a collection device and collection state information of the collection device; determining feature matching data of the road scene data, determining three-dimensional feature data of road markers on the target road based on the road scene data and the feature matching data; the feature matching data comprises at least one common view matching pair; the image data included in the common view matching pair has similar or identical image content; constructing a road surface model of the target road based on the three-dimensional feature data and the collection state information.

2. The method of claim 1, wherein, The determination of the feature matching data of the road scene data comprises: performing feature extraction on the image data in the road scene data to obtain pixel feature information of the image data; determining the feature matching data of the road scene data based on the pixel feature information.

3. The method of claim 1, wherein, The determination of the three-dimensional feature data of the road markers on the target road based on the road scene data and the feature matching data comprises: based on the feature matching data and the collection state information of the plurality of collection devices, performing three-dimensional feature reconstruction on the road scene data to obtain a plurality of three-dimensional road data; aligning the plurality of three-dimensional road data to obtain the three-dimensional feature data.

4. The method of claim 3, wherein, The three-dimensional feature reconstruction of the road scene data based on the feature matching data and the collection state information of each of the plurality of collection devices to obtain a plurality of three-dimensional road data comprises: based on the extrinsic parameters of each of the plurality of collection devices, correcting the collection state information of each of the plurality of collection devices to obtain corrected collection state information of each of the plurality of collection devices; based on the feature matching data and the corrected collection state information of each of the plurality of collection devices, performing three-dimensional feature reconstruction on the road scene data to obtain the plurality of three-dimensional road data.

5. The method of claim 3, wherein, The alignment of the plurality of three-dimensional road data to obtain the three-dimensional feature data comprises: based on the corrected collection state information of each of the plurality of collection devices, determining shared field of view information of the road scene data; based on the shared field of view information, aligning the plurality of three-dimensional road data to obtain the three-dimensional feature data.

6. The method according to any one of claims 1-5, characterized in that, The construction of the road surface model of the target road based on the three-dimensional feature data and the collection state information comprises: based on the three-dimensional feature data, updating the collection state information of each of the plurality of collection devices to obtain updated collection state information of each of the plurality of collection devices; based on the updated collection state information of each of the plurality of collection devices, performing semantic recognition on the three-dimensional feature data to determine pixel distribution of road markers on the target road in a two-dimensional image space; based on the pixel distribution of the road markers on the target road in the two-dimensional image space, rendering the road markers on the target road to obtain the road surface model of the target road.

7. The method of claim 6, wherein, The method further comprises: based on the road scene data, determining feature data of a dynamic target on the target road; Render the feature data of the dynamic target into the road surface model to obtain a three-dimensional dynamic model corresponding to the dynamic target.

8. The method of claim 7, wherein, The collection state information includes a pose of the collection device when collecting image data, and internal parameter data and external parameter data of the collection device.

9. A road surface model construction device characterized by comprising: The device comprises an acquisition unit, a determination unit and a construction unit. The acquisition unit is configured to acquire a plurality of road scene data of a target road; the road scene data comprises image data collected by a collection device, and collection state information of the collection device. The determination unit is configured to determine three-dimensional feature data of road markers on the target road based on the road scene data and feature matching data of the road scene data; the feature matching data comprises a plurality of common view matching pairs; the common view matching pairs comprise a plurality of image data having similar or identical image content. The construction unit is configured to construct a road surface model of the target road based on the three-dimensional feature data and the collection state information.

10. A pavement model building system, characterized by, The system comprises a computing resource scheduler, a data resource manager and a plurality of processors. The computing resource scheduler is configured to acquire a plurality of road scene data of a target road from the data resource manager; the road scene data comprises a plurality of image data collected by a collection device, and collection state information of the collection device. The computing resource scheduler is further configured to determine a target processor according to resource usage of each processor, and send the plurality of road scene data to the target processor; wherein the target processor is a processor in the plurality of processors whose resource occupancy rate is less than a preset threshold. The target processor is configured to: determine three-dimensional feature data of road markers on the target road based on the plurality of road scene data and feature matching data of the plurality of road scene data; the feature matching data comprises a plurality of common view matching pairs; the common view matching pairs comprise a plurality of image data having similar or identical image content; construct a road surface model of the target road based on the three-dimensional feature data and the collection state information of the collection device.

11. An electronic device, comprising: comprise: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the method of any one of claims 1 to 8.

12. A computer-readable storage medium, characterized in that, When the computer-executable instructions stored in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can perform the method of any one of claims 1 to 8.

13. A computer program product comprising instructions, characterized in that, When the instructions are run by a computer, the computer is caused to perform the method of any one of claims 1 to 8.

Images