High-precision map optimization method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of automatic driving, and relates to a high-precision map optimization method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring original map related data from at least one data source; performing preprocessing operation on the original map related data, wherein the preprocessing operation comprises cleaning and denoising and space-time alignment; performing feature matching among different data sources according to the pre-processed map related data to obtain a successfully matched feature set; according to the successfully matched feature set, training to obtain a target map model; performing space optimization on the target map model to obtain the target map model after version updating; and pushing the target map model of which the version is updated to a vehicle system through the cloud. By adopting the method, dynamic updating and real-time synchronization can be realized, and the accuracy and timeliness of map information are ensured.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a high-precision map optimization method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the development of autonomous driving technology, high-precision maps have become one of the key technologies to achieve autonomous driving. High-precision maps can provide information with centimeter-level accuracy, such as lane information and surrounding environment details. However, information such as road conditions, traffic signs, and construction areas often change, and keeping maps highly accurate and up-to-date is a huge challenge. The main challenges facing high-precision maps currently include the timeliness of map data, accurate positioning in complex environments, and how to efficiently process and integrate massive data from multiple sensors.

[0003] Therefore, there is an urgent need for a high-precision map optimization method, device, computer equipment, computer-readable storage medium and computer program product that can achieve dynamic updating and real-time synchronization to ensure the accuracy and timeliness of map information. Summary of the invention

[0004] Based on this, it is necessary to provide a high-precision map optimization method, device, computer equipment, computer-readable storage medium and computer program product that can achieve dynamic updating and real-time synchronization to ensure the accuracy and timeliness of map information in response to the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a high-precision map optimization method, comprising:

[0006] collecting raw map-related data from at least one data source;

[0007] Performing a preprocessing operation on the original map-related data, wherein the preprocessing operation includes cleaning, denoising and time-space alignment;

[0008] According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set;

[0009] According to the successfully matched feature set, the target map model is trained and obtained;

[0010] Performing spatial optimization on the target map model to obtain an updated version of the target map model;

[0011] The updated version of the target map model is pushed to the vehicle system via the cloud.

[0012] In one embodiment, before training to obtain a target map model based on the successfully matched feature set, the method further includes:

[0013] Use probabilistic statistical methods and machine learning models to perform data fusion on feature sets;

[0014] Evaluate the data quality of the fused feature set;

[0015] When the data quality does not meet the preset mapping standards, the fused feature set is further optimized by using path smoothing and topological structure.

[0016] In one embodiment, after pushing the updated version of the target map model to the vehicle system via the cloud, the method further includes:

[0017] Collect real-time feedback data of the vehicle driving environment;

[0018] The performance of the target map model is evaluated according to the real-time feedback data of the vehicle driving environment, and the target map model is spatially optimized.

[0019] In one embodiment, the collecting of real-time feedback data of the vehicle driving environment includes:

[0020] Identify the vehicle's driving environment and collect real-time feedback data of the vehicle's driving environment;

[0021] The real-time feedback data is connected to the vehicle system through an API interface and formatted to be compatible with the internal data format of the vehicle system.

[0022] In one embodiment, the collecting of real-time feedback data of the vehicle driving environment includes:

[0023] Using laser radar to collect a laser point cloud data set of the driving environment, performing cluster analysis on the laser point cloud data set, and identifying road edges and isolation zones according to the analysis results;

[0024] A camera device is used to collect environmental image data of the driving environment, and based on the environmental image data, a deep learning algorithm is used to identify traffic signs and lane lines.

[0025] In one embodiment, evaluating the performance of the target map model according to the real-time feedback data of the vehicle driving environment and performing spatial optimization on the target map model includes:

[0026] Performing spatiotemporal alignment of the laser point cloud data set and the environmental image data using the timestamp and location information recorded by the vehicle system;

[0027] Using a feature matching algorithm, the laser point cloud data set and the environmental image data are matched, and the performance of the target map model is evaluated according to the matching result;

[0028] When the performance of the target map model is lower than a preset performance threshold, the accuracy of the target map model is optimized by combining the GPS trajectory and the vehicle communication system V2X.

[0029] In a second aspect, the present application also provides a high-precision map optimization device, comprising:

[0030] A data collection module, used for collecting original map-related data from at least one data source;

[0031] A preprocessing module, used for performing preprocessing operations on the original map-related data, wherein the preprocessing operations include cleaning, denoising and spatiotemporal alignment;

[0032] A matching module is used to perform feature matching between different data sources based on the preprocessed map-related data to obtain a successfully matched feature set;

[0033] The model training module is used to train and obtain the target map model according to the successfully matched feature set;

[0034] A model optimization module, used for performing spatial optimization on the target map model to obtain an updated version of the target map model;

[0035] The push module is used to push the updated version of the target map model to the vehicle system through the cloud.

[0036] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0037] collecting raw map-related data from at least one data source;

[0038] Performing a preprocessing operation on the original map-related data, wherein the preprocessing operation includes cleaning, denoising and time-space alignment;

[0039] According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set;

[0040] According to the successfully matched feature set, the target map model is trained and obtained;

[0041] Performing spatial optimization on the target map model to obtain an updated version of the target map model;

[0042] The updated version of the target map model is pushed to the vehicle system via the cloud.

[0043] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0044] collecting raw map-related data from at least one data source;

[0045] Performing a preprocessing operation on the original map-related data, wherein the preprocessing operation includes cleaning, denoising and time-space alignment;

[0046] According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set;

[0047] According to the successfully matched feature set, the target map model is trained and obtained;

[0048] Performing spatial optimization on the target map model to obtain an updated version of the target map model;

[0049] The updated version of the target map model is pushed to the vehicle system via the cloud.

[0050] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0051] collecting raw map-related data from at least one data source;

[0052] Performing a preprocessing operation on the original map-related data, wherein the preprocessing operation includes cleaning, denoising and time-space alignment;

[0053] According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set;

[0054] According to the successfully matched feature set, the target map model is trained and obtained;

[0055] Performing spatial optimization on the target map model to obtain an updated version of the target map model;

[0056] The updated version of the target map model is pushed to the vehicle system via the cloud.

[0057] The above-mentioned high-precision map optimization method, device, computer equipment, computer-readable storage medium and computer program product can ensure the quality of map data and improve the accuracy and reliability of the map by collecting original map data from different data sources and performing preprocessing operations such as cleaning, denoising and spatiotemporal alignment. By matching the features between different data sources, common information in each data source can be discovered and integrated, thereby enhancing the consistency and integrity of the map data. By training to obtain the target map model, a map model with better performance can be constructed, which can more accurately reflect the geographic information of the real world. Spatial optimization of the target map model can improve the storage efficiency and rendering speed of map data and enhance the user experience of the map. By pushing an updated version of the map model to the vehicle system through the cloud, real-time updating of map data can be achieved to ensure that the map information used by users is the latest. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0059] FIG1 is a diagram of an application environment of a high-precision map optimization method in one embodiment;

[0060] FIG2 is a schematic diagram of a process flow of a high-precision map optimization method in one embodiment;

[0061] FIG3 is a schematic diagram of a flow chart of a high-precision map optimization method in another embodiment;

[0062] FIG4 is a structural block diagram of a high-precision map optimization device in one embodiment;

[0063] FIG. 5 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0065] The high-precision map optimization method provided in the embodiment of the present application can be applied to the application environment shown in Figure 1. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers.

[0066] The server 104 is used to collect original map-related data from at least one data source; perform preprocessing operations on the original map-related data, the preprocessing operations including cleaning, denoising and spatiotemporal alignment; perform feature matching between different data sources based on the preprocessed map-related data to obtain a successfully matched feature set; train and obtain a target map model based on the successfully matched feature set; perform spatial optimization on the target map model to obtain an updated version of the target map model; and push the updated version of the target map model to the vehicle system of the terminal 102 via the cloud.

[0067] Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0068] In an exemplary embodiment, as shown in FIG2 , a high-precision map optimization method is provided, which is described by taking the method applied to the server in FIG1 as an example, and includes the following steps S202 to S212. Among them:

[0069] Step S202: collecting original map-related data from at least one data source.

[0070] Specifically, at least one data source includes vehicle-mounted sensor data, satellite images, drone mapping data, and V2X communication data, wherein:

[0071] On-board sensor data: This includes data collected using on-board sensors such as LiDAR, cameras, radar, etc. These sensors can provide real-time, high-resolution three-dimensional information of the vehicle's surroundings, including road edges, traffic signs, obstacles, etc.

[0072] Satellite imagery: High-resolution images acquired by satellites can provide static background information such as macro terrain, building layout, etc. These data are often used to provide macro context and reference frame for maps.

[0073] Drone mapping data: Drones can provide detailed aerial photography and scanning of specific areas, collecting high-precision data on ground features and structures, especially in areas that are difficult to access using vehicle-mounted sensors or satellite imagery.

[0074] V2X communication data: V2X (Vehicle-to-Everything) communication technology allows vehicles to exchange data with other vehicles, infrastructure, pedestrians, etc. This data includes real-time information such as road conditions, traffic flow, accidents and construction areas.

[0075] When collecting this data, the system needs to be able to process and integrate data of various formats and precisions. This requires the system to be highly flexible and compatible, able to identify and integrate data from different sources. In addition, the data collection process also needs to take into account the timeliness and accuracy of the data to ensure that the map information is updated in real time and is highly accurate.

[0076] Step S204, preprocessing the original map-related data, including cleaning, denoising and spatiotemporal alignment.

[0077] Specifically, in the process of building high-precision maps, the original map data collected from various data sources often contains noise, errors or inconsistencies, which will affect the accuracy and efficiency of subsequent data processing. Therefore, preprocessing operations are a key step to ensure data quality. The following is a detailed description of this process:

[0078] Data cleaning involves identifying and removing invalid or erroneous information from the data. For example, onboard sensors can produce noisy data, such as abnormal altitude readings or erroneous distance measurements, due to technical limitations or environmental interference. The data cleaning process will use algorithms to identify these outliers and remove or replace them from the dataset to reduce the impact on the final map accuracy.

[0079] Denoising is a subset of data cleaning that focuses on reducing random noise in the data. This includes using filtering algorithms, such as Gaussian filtering or median filtering, to smooth the data and remove abnormal fluctuations. For example, the point cloud in LiDAR data will contain noisy points due to reflectivity changes or multipath effects. The denoising process will help remove these inaccurate points, thereby improving the reliability of the data.

[0080] Spatiotemporal alignment: Since data comes from different sensors and data sources, they are captured at different timestamps and spatial coordinate systems. The spatiotemporal alignment step ensures that all data are compared and integrated under the same time frame and spatial reference. This includes converting all data to a unified coordinate system (such as WGS84) and synchronizing them according to their capture time. This step is critical to ensure the consistency and accuracy of map data, especially in dynamically changing environments such as urban transportation networks.

[0081] Step S206: performing feature matching between different data sources based on the preprocessed map-related data to obtain a successfully matched feature set.

[0082] Specifically, after data preprocessing, key features need to be extracted from various data sources first. These features include road edges, traffic signs, lane lines, building outlines, etc. Feature extraction is usually implemented using computer vision and deep learning techniques to ensure the accuracy and reliability of the features. Feature matching is the comparison of features extracted from different data sources to identify and associate features representing the same entity or geographic element. This step involves complex algorithms such as feature point matching algorithms (such as ICP-IterativeClosestPoint algorithm), descriptor-based matching algorithms (such as SIFT, SURF, or ORB algorithms), or using deep learning models to identify and match features.

[0083] Step S208: training and obtaining a target map model based on the successfully matched feature set.

[0084] Specifically, once the features are matched, the data is used to train or optimize the target map model. In the field of machine learning, training usually refers to using a large amount of data to adjust the parameters of the model until the model can accurately perform a specific task. In this context, the task is to create an accurate, high-precision map that reflects current road conditions.

[0085] Step S210, spatially optimizing the target map model to obtain an updated version of the target map model.

[0086] Specifically, spatial optimization refers to improving the geometric and spatial information in the map model to improve the precision and accuracy of the map. This includes adjusting parameters such as the shape, width, slope, curvature, etc. of the road to more accurately reflect the actual situation. Spatial optimization involves complex algorithms such as 3D modeling and geometric correction.

[0087] During the spatial optimization process, information from multiple sensors and data sources needs to be fused together. For example, LiDAR point cloud data, camera images, GPS trajectories, and V2X (vehicle-to-everything) data are combined to obtain a more comprehensive and accurate understanding of the environment. Features in the map model, such as lane lines, traffic signs, road signs, etc., will change due to road construction, traffic accidents or other events. During the spatial optimization process, these changes need to be identified and updated to the map model. During the optimization process, the performance of the map model needs to be continuously evaluated to ensure the effectiveness of the optimization measures. This involves using specialized software tools to test the accuracy, completeness, and consistency of the map. After spatial optimization, the map model will be improved to generate a new, more accurate version. This updated version will be used to replace or update the existing map data to ensure that the autonomous driving system and other applications can use the latest and most accurate map information.

[0088] In the field of autonomous driving, road conditions change rapidly. Therefore, spatial optimization and updating is an ongoing process that needs to be performed regularly or in real time to keep the map data up to date.

[0089] Step S212: Push the updated version of the target map model to the vehicle system via the cloud.

[0090] Specifically, the processed data is used to update the target map model and generate a new map version that contains the latest road conditions, traffic signs, construction areas and other information. The updated map model is uploaded to the cloud server and can be pushed to the navigation system of the autonomous driving vehicle via a wireless network (such as 4G / 5G). The vehicle receives the updated map data and applies it to the autonomous driving system to ensure that the vehicle's perception, positioning, decision-making and control modules can operate based on the latest and most accurate map information.

[0091] This cloud push mechanism ensures that autonomous vehicles can respond quickly to road changes, improving the safety and reliability of autonomous driving. At the same time, it also reduces the vehicle's dependence on local computing resources, allowing the vehicle to focus more on real-time decision-making and control tasks. In addition, cloud push also supports centralized management and rapid distribution of map data, improving the efficiency and coverage of map updates.

[0092] In the above-mentioned high-precision map optimization method, by collecting original map data from different data sources and performing preprocessing operations such as cleaning, denoising, and spatiotemporal alignment, the quality of map data can be ensured and the accuracy and reliability of the map can be improved. By matching the features between different data sources, common information in each data source can be discovered and integrated, thereby enhancing the consistency and integrity of map data. By training to obtain the target map model, a map model with better performance can be built that can more accurately reflect the geographic information of the real world. Spatial optimization of the target map model can improve the storage efficiency and rendering speed of map data and enhance the user experience of the map. By pushing an updated version of the map model to the vehicle system through the cloud, real-time updating of map data can be achieved to ensure that the map information used by users is the latest.

[0093] In an exemplary embodiment, as shown in FIG3 , before training to obtain a target map model based on a successfully matched feature set, the method further includes:

[0094] Step S302, using probability statistics method and machine learning model to perform data fusion on the feature set;

[0095] Step S304, evaluating the data quality of the fused feature set;

[0096] Step S306: When the data quality does not meet the preset mapping standard, the fused feature set is further optimized by using path smoothing and topological structure.

[0097] Specifically, after collecting various sensor data (such as LiDAR, camera, GPS, etc.), the relevant features in these data sets need to be matched and aligned first. Then, these matched features are fused using probabilistic statistical methods and machine learning models. The purpose of data fusion is to integrate information from different sensors to obtain a more accurate and comprehensive feature representation.

[0098] For example, algorithms such as Kalman filtering, particle filtering, or deep learning models such as convolutional neural networks (CNNs) can be used to fuse image and point cloud data.

[0099] After the fusion is completed, the data quality of the result needs to be evaluated to ensure that the fused data meets the accuracy and completeness requirements for mapping. The evaluation includes checking the integrity, consistency, accuracy and timeliness of the data. For example, it can be done by calculating the error range, comparing the consistency of different data sources, or using certain quality assessment indicators.

[0100] If the evaluation results show that the quality of the fused feature set data does not meet the preset mapping standards, further optimization is required. For the vehicle trajectory data, smoothing algorithms (such as Gaussian filtering, spline interpolation, etc.) can be used to reduce noise and errors to make the path smoother and more realistic.

[0101] For the topological relationships of the map, such as road connectivity and intersection structure, it is necessary to ensure that they are correct. This involves adjusting and optimizing the topological structure of complex areas such as intersections and road forks to ensure the logical consistency of the map.

[0102] In this embodiment, during the mapping process, the above operation methods can ensure that the map model accurately reflects the geometric and semantic information of the real world, providing the necessary environmental cognition capabilities for the autonomous driving vehicle.

[0103] In an exemplary embodiment, after the updated version of the target map model is pushed to the vehicle system via the cloud, the following steps are further included:

[0104] Collect real-time feedback data of the vehicle driving environment;

[0105] Based on the real-time feedback data of the vehicle driving environment, the performance of the target map model is evaluated and the target map model is spatially optimized.

[0106] Specifically, when an autonomous vehicle is driving, its various sensors (such as cameras, lidar, GPS, etc.) will continuously collect real-time data about the surrounding environment. This data includes information such as the location of lane lines, the status of traffic signs, road conditions, and traffic flow, which is crucial for detecting the accuracy and completeness of the map.

[0107] The collected real-time feedback data needs to be compared and analyzed with the current high-precision map to evaluate the performance of the map. The evaluation includes checking the accuracy and timeliness of the map information and whether it meets the needs of the autonomous driving system. Through this evaluation, errors or outdated information in the map can be found, such as route changes caused by road construction, new traffic signs, etc. Based on the evaluation results, the map model needs to be spatially optimized to update and improve the map data.

[0108] Spatial optimization includes adjusting road geometry models, updating traffic signs and signals, and redrawing lane lines to ensure that map information is consistent with actual road conditions. This process involves complex algorithms such as machine learning, computer vision, and spatial analysis to automatically process and update map data. The optimized map model needs to be pushed to the vehicle system again, and then real-time feedback data is collected again for evaluation. Through continuous iteration, it can be ensured that high-precision maps are always kept up to date, providing continuous and reliable environmental information for autonomous driving.

[0109] In this embodiment, by maintaining the high-precision map, the map data is continuously updated and optimized to adapt to the ever-changing road conditions. This dynamic update mechanism is crucial for the safe and efficient operation of autonomous vehicles.

[0110] In an exemplary embodiment, collecting real-time feedback data of the vehicle driving environment includes:

[0111] Identify the vehicle's driving environment and collect real-time feedback data of the vehicle's driving environment;

[0112] The real-time feedback data is connected to the vehicle system through the API interface and the format is converted to be compatible with the internal data format of the vehicle system.

[0113] Specifically, autonomous vehicles use various sensors (such as cameras, lidar, radar, GPS, inertial measurement unit (IMU), etc.) to perceive the surrounding driving environment. These sensors can detect objects near the vehicle, road signs, traffic signals, lane lines, road conditions, traffic flow and other information, providing the vehicle with 360-degree environmental perception capabilities. Sensors monitor and record environmental changes around the vehicle in real time, generating a large amount of raw data. These data include images, point clouds, signal strength, location coordinates, speed and acceleration, etc., which are the basis for autonomous driving systems to make decisions and control.

[0114] In order to effectively integrate the collected raw data into the vehicle's autonomous driving system, the data needs to be transferred to the vehicle's internal system through an application programming interface (API). API provides a standardized method that allows data exchange and communication between different devices and software. Since the data collected by sensors has different formats and standards, and the vehicle system has its own data format requirements, the raw data needs to be converted into a format that the system can understand and process, which involves operations such as data parsing, cleaning, normalization, and packaging to ensure that the data can be correctly received, parsed, and used by the vehicle system.

[0115] In this embodiment, this process ensures that the autonomous driving vehicle can collect and process information about the surrounding environment in real time, providing necessary data support for the vehicle's positioning, navigation, decision-making and control. In this way, the autonomous driving system can respond to environmental changes in a timely manner and improve driving safety and efficiency.

[0116] In an exemplary embodiment, collecting real-time feedback data of the vehicle driving environment includes:

[0117] Use laser radar to collect laser point cloud data sets of the driving environment, perform cluster analysis on the laser point cloud data sets, and identify road edges and isolation zones based on the analysis results;

[0118] A camera device is used to collect environmental image data of the driving environment, and based on the environmental image data, a deep learning algorithm is used to identify traffic signs and lane lines.

[0119] Specifically, LiDAR measures the distance of an object by emitting laser pulses and receiving the reflected light, generating a three-dimensional data set containing a large number of points, namely a point cloud. These point cloud data contain the spatial position information of objects in the driving environment and are an important data source for autonomous vehicles to perceive the environment. Cluster analysis is a method of dividing the points in a data set into multiple groups (or "clusters") based on similarity, so that the points in the same cluster are as similar as possible, and the points between different clusters are as different as possible. In autonomous driving, clustering algorithms are often used to identify different objects from laser point cloud data, such as vehicles, pedestrians, road edges, and isolation belts.

[0120] Cluster analysis can distinguish point cloud data belonging to the edge of the road and the median strip. This information is crucial for autonomous vehicles to understand their location and road structure, helping them stay in the correct lane and avoid potential obstacles.

[0121] Cameras are another important sensor that can capture image data of the road environment. These image data contain rich color and texture information, which plays an important role in identifying traffic signs, lane lines, etc. Deep learning algorithms, especially convolutional neural networks (CNNs), have performed well in the field of image recognition. Through training, deep learning models can automatically learn and extract features of traffic signs and lane lines from images to achieve high-precision recognition. For example, CNN can be used to classify traffic signs with an accuracy of 96%. In addition, deep learning is also widely used in lane line detection, and accurate lane line recognition is achieved through methods such as semantic segmentation, detection, and parametric curve fitting.

[0122] In this embodiment, these steps together constitute the process of real-time feedback data collection and processing of the autonomous driving vehicle, providing the vehicle with the necessary environmental perception information to achieve safe and accurate driving decisions.

[0123] In an exemplary embodiment, the performance of the target map model is evaluated according to the real-time feedback data of the vehicle driving environment, and the target map model is spatially optimized, including:

[0124] The laser point cloud data set and the environmental image data are temporally and spatially aligned using the timestamp and location information recorded by the vehicle system;

[0125] Use feature matching algorithms to match the laser point cloud data set and the environmental image data, and evaluate the performance of the target map model based on the matching results;

[0126] When the performance of the target map model is lower than the preset performance threshold, the accuracy of the target map model is optimized by combining the GPS trajectory and the vehicle communication system V2X.

[0127] Specifically, first, the laser point cloud dataset and the environmental image data need to be aligned in time and space through the timestamps and location information recorded by the vehicle system. This means integrating data from different sources (such as the point cloud data of the lidar and the image data of the camera) into the same coordinate system and time frame for effective comparison and analysis. This alignment is the basis for subsequent analysis and ensures the consistency and comparability of the data. The multi-target tracking algorithm with spatiotemporal feature alignment ensures the temporal consistency of the historical frame information and the current frame information through the feature alignment module to reduce the false detection rate.

[0128] Next, the laser point cloud dataset and the environmental image data are matched using a feature matching algorithm. Feature matching refers to finding the same or similar feature points between different data sources so that these data can be associated and compared. Based on the matching results, the performance of the target map model can be evaluated. This step involves complex algorithms such as feature extraction and matching techniques based on deep learning.

[0129] Based on the results of feature matching, the performance of the target map model is evaluated. This includes checking the map's accuracy, completeness, and whether it can accurately reflect actual road conditions. Performance evaluation is a prerequisite for optimizing the map model, because only by understanding the shortcomings of the current model can targeted improvements be made.

[0130] If the performance of the target map model is lower than the preset performance threshold, it needs to be spatially optimized. This involves combining GPS tracks and the vehicle communication system V2X (Vehicle-to-Everything) to optimize the accuracy of the target map model. Spatial optimization includes adjusting various elements on the map, such as the location, shape, and attributes of the road, to ensure the accuracy and real-time nature of the map data. In this way, the availability and reliability of the map can be improved, thereby providing better support for autonomous driving.

[0131] In this embodiment, by continuously collecting, processing and analyzing real-time data, it is possible to ensure that the map model is always in the latest and most accurate state.

[0132] The most detailed embodiment of this application is:

[0133] Scenario setting: Assume that an autonomous vehicle is driving in an urban environment and needs to update its high-precision map in real time to cope with the dynamically changing traffic environment.

[0134] Implementation steps:

[0135] 1. Data collection and preprocessing:

[0136] Onboard LiDAR, cameras, and GPS / IMU sensors collect data about the surrounding environment.

[0137] V2X communication systems receive data from other vehicles and road infrastructure.

[0138] Through the data access interface, all data are collected centrally and initially cleaned and formatted to remove invalid or erroneous information.

[0139] 2. Feature extraction:

[0140] Use LiDAR point cloud data for cluster analysis to identify key features such as road edges and median strips.

[0141] Use deep learning algorithms to process camera images, identify traffic signs and lane markings, and extract color and texture features.

[0142] GPS / IMU data is used for vehicle positioning and motion trajectory tracking.

[0143] 3. Multi-source data fusion:

[0144] Through timestamp and location information, LiDAR, camera, GPS / IMU and V2X data are aligned in the same space-time framework.

[0145] Use feature matching algorithms, such as the ICP algorithm, to match LiDAR and camera data, and combine GPS trajectories to optimize the accuracy of the road model.

[0146] V2X information is integrated with static map data to achieve real-time updates of road conditions.

[0147] 4. Heterogeneous data fusion and consistency verification:

[0148] Using algorithms such as Kalman filtering, Bayesian networks or deep learning models, the uncertainty and reliability of the data are comprehensively considered to output the most likely scenario description.

[0149] Resolve data conflicts through intelligent algorithms to ensure the consistency and integrity of map data.

[0150] 5. Spatial data optimization and reconstruction:

[0151] Further spatial optimization is performed on the fused data, such as smoothing and gridding, to improve the accuracy and usability of the map.

[0152] Dynamically adjust the accuracy and detail of map display according to vehicle driving status and environmental complexity.

[0153] 6. Real-time map update and release:

[0154] The optimized data is compiled into the latest map version and quickly pushed to autonomous driving vehicles through the cloud, supporting both online and offline modes to achieve real-time updates of map data.

[0155] The update frequency is flexibly adjusted according to the speed and importance of road condition changes to ensure that the vehicle has the latest and most accurate driving environment information.

[0156] 7. Feedback and optimization loop:

[0157] The system continuously collects feedback information from the vehicle and the environment, such as positioning accuracy, path planning effectiveness, etc., to evaluate map performance.

[0158] Based on feedback information, the data fusion algorithm and the performance of the entire system are continuously adjusted and optimized to form a closed-loop self-improvement mechanism.

[0159] Application example effect: In the implementation example, the autonomous vehicle identified and avoided lane closures caused by temporary construction in advance based on the real-time updated high-precision map, and successfully detoured to an alternative route. At complex intersections, by integrating camera and V2X data, the vehicle accurately judged the status of traffic lights and pedestrian crossings, and made safe driving decisions. This implementation example shows how to improve the safety and efficiency of autonomous vehicles through multi-source data fusion and real-time dynamic update of an adaptive high-precision map system.

[0160] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0161] Based on the same inventive concept, the embodiment of the present application also provides a high-precision map optimization device for implementing the high-precision map optimization method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more high-precision map optimization device embodiments provided below can refer to the limitations on the high-precision map optimization method above, and will not be repeated here.

[0162] In an exemplary embodiment, as shown in FIG4 , a high-precision map optimization device is provided, comprising:

[0163] A data collection module 402, used to collect original map-related data from at least one data source;

[0164] A preprocessing module 404 is used to perform preprocessing operations on the original map-related data, and the preprocessing operations include cleaning, denoising and time-space alignment;

[0165] A matching module 406 is used to perform feature matching between different data sources based on the pre-processed map-related data to obtain a successfully matched feature set;

[0166] A model training module 408 is used to train and obtain a target map model according to the successfully matched feature set;

[0167] The model optimization module 410 is used to perform spatial optimization on the target map model to obtain an updated version of the target map model;

[0168] The push module 412 is used to push the updated version of the target map model to the vehicle system via the cloud.

[0169] In an exemplary embodiment, the data fusion module is used to perform data fusion on the feature set using a probabilistic statistical method and a machine learning model;

[0170] Data quality assessment module, used to assess the data quality of the fused feature set;

[0171] The data optimization module is used to further optimize the fused feature set by using path smoothing and topological structure when the data quality does not meet the preset mapping standard.

[0172] In an exemplary embodiment, the data collection module 402 is also used to collect real-time feedback data of the vehicle driving environment;

[0173] The model optimization module 410 is also used to evaluate the performance of the target map model according to the real-time feedback data of the vehicle driving environment, and to perform spatial optimization on the target map model.

[0174] In an exemplary embodiment, the data acquisition module 402 is also used to identify the driving environment of the vehicle and collect real-time feedback data of the vehicle driving environment; the preprocessing module 404 is also used to connect the real-time feedback data to the vehicle system through the API interface and perform format conversion to be compatible with the internal data format of the vehicle system.

[0175] In an exemplary embodiment, the data acquisition module 402 is further used to collect a laser point cloud data set of the driving environment using a laser radar;

[0176] Feature extraction module, used to perform cluster analysis on the laser point cloud data set and identify road edges and isolation zones based on the analysis results;

[0177] The data acquisition module 402 is also used to collect environmental image data of the driving environment using a camera device;

[0178] The feature extraction module is also used to identify traffic signs and lane lines based on environmental image data using deep learning algorithms.

[0179] In an exemplary embodiment, the model optimization module 410 is also used to align the laser point cloud dataset and the environmental image data in time and space through the timestamp and location information recorded by the vehicle system; use a feature matching algorithm to match the laser point cloud dataset and the environmental image data, and evaluate the performance of the target map model based on the matching results; when the performance of the target map model is lower than a preset performance threshold, combine the GPS trajectory and the vehicle communication system V2X to optimize the accuracy of the target map model.

[0180] Each module in the above-mentioned high-precision map optimization device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0181] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG5. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store original map-related data collected from at least one data source. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a high-precision map optimization method is implemented.

[0182] Those skilled in the art will understand that the structure shown in FIG. 5 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0183] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0184] collecting raw map-related data from at least one data source;

[0185] Preprocessing the original map-related data, including cleaning, denoising and spatiotemporal alignment;

[0186] According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set;

[0187] According to the successfully matched feature set, the target map model is trained and obtained;

[0188] Perform spatial optimization on the target map model to obtain an updated version of the target map model;

[0189] The updated version of the target map model is pushed to the vehicle system via the cloud.

[0190] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0191] Use probabilistic statistical methods and machine learning models to perform data fusion on feature sets;

[0192] Evaluate the data quality of the fused feature set;

[0193] When the data quality does not meet the preset mapping standards, the fused feature set is further optimized by using path smoothing and topological structure.

[0194] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0195] Collect real-time feedback data of the vehicle driving environment;

[0196] Based on the real-time feedback data of the vehicle driving environment, the performance of the target map model is evaluated and the target map model is spatially optimized.

[0197] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0198] Identify the vehicle's driving environment and collect real-time feedback data of the vehicle's driving environment;

[0199] The real-time feedback data is connected to the vehicle system through the API interface and the format is converted to be compatible with the internal data format of the vehicle system.

[0200] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0201] Use laser radar to collect laser point cloud data sets of the driving environment, perform cluster analysis on the laser point cloud data sets, and identify road edges and isolation zones based on the analysis results;

[0202] A camera device is used to collect environmental image data of the driving environment, and based on the environmental image data, a deep learning algorithm is used to identify traffic signs and lane lines.

[0203] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0204] The laser point cloud data set and the environmental image data are temporally and spatially aligned using the timestamp and location information recorded by the vehicle system;

[0205] Use feature matching algorithms to match the laser point cloud data set and the environmental image data, and evaluate the performance of the target map model based on the matching results;

[0206] When the performance of the target map model is lower than the preset performance threshold, the accuracy of the target map model is optimized by combining the GPS trajectory and the vehicle communication system V2X.

[0207] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0208] collecting raw map-related data from at least one data source;

[0209] Preprocessing the original map-related data, including cleaning, denoising and spatiotemporal alignment;

[0210] According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set;

[0211] According to the successfully matched feature set, the target map model is trained and obtained;

[0212] Perform spatial optimization on the target map model to obtain an updated version of the target map model;

[0213] The updated version of the target map model is pushed to the vehicle system via the cloud.

[0214] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0215] Use probabilistic statistical methods and machine learning models to perform data fusion on feature sets;

[0216] Evaluate the data quality of the fused feature set;

[0217] When the data quality does not meet the preset mapping standards, the fused feature set is further optimized by using path smoothing and topological structure.

[0218] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0219] Collect real-time feedback data of the vehicle driving environment;

[0220] Based on the real-time feedback data of the vehicle driving environment, the performance of the target map model is evaluated and the target map model is spatially optimized.

[0221] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0222] Identify the vehicle's driving environment and collect real-time feedback data of the vehicle's driving environment;

[0223] The real-time feedback data is connected to the vehicle system through the API interface and the format is converted to be compatible with the internal data format of the vehicle system.

[0224] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0225] Use laser radar to collect laser point cloud data sets of the driving environment, perform cluster analysis on the laser point cloud data sets, and identify road edges and isolation zones based on the analysis results;

[0226] A camera device is used to collect environmental image data of the driving environment, and based on the environmental image data, a deep learning algorithm is used to identify traffic signs and lane lines.

[0227] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0228] The laser point cloud data set and the environmental image data are temporally and spatially aligned using the timestamp and location information recorded by the vehicle system;

[0229] Use feature matching algorithms to match the laser point cloud data set and the environmental image data, and evaluate the performance of the target map model based on the matching results;

[0230] When the performance of the target map model is lower than the preset performance threshold, the accuracy of the target map model is optimized by combining the GPS trajectory and the vehicle communication system V2X.

[0231] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0232] collecting raw map-related data from at least one data source;

[0233] Preprocessing the original map-related data, including cleaning, denoising and spatiotemporal alignment;

[0234] According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set;

[0235] According to the successfully matched feature set, the target map model is trained and obtained;

[0236] Perform spatial optimization on the target map model to obtain an updated version of the target map model;

[0237] The updated version of the target map model is pushed to the vehicle system via the cloud.

[0238] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0239] Use probabilistic statistical methods and machine learning models to perform data fusion on feature sets;

[0240] Evaluate the data quality of the fused feature set;

[0241] When the data quality does not meet the preset mapping standards, the fused feature set is further optimized by using path smoothing and topological structure.

[0242] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0243] Collect real-time feedback data of the vehicle driving environment;

[0244] Based on the real-time feedback data of the vehicle driving environment, the performance of the target map model is evaluated and the target map model is spatially optimized.

[0245] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0246] Identify the vehicle's driving environment and collect real-time feedback data of the vehicle's driving environment;

[0247] The real-time feedback data is connected to the vehicle system through the API interface and the format is converted to be compatible with the internal data format of the vehicle system.

[0248] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0249] Use laser radar to collect laser point cloud data sets of the driving environment, perform cluster analysis on the laser point cloud data sets, and identify road edges and isolation zones based on the analysis results;

[0250] A camera device is used to collect environmental image data of the driving environment, and based on the environmental image data, a deep learning algorithm is used to identify traffic signs and lane lines.

[0251] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0252] The laser point cloud data set and the environmental image data are temporally and spatially aligned using the timestamp and location information recorded by the vehicle system;

[0253] Use feature matching algorithms to match the laser point cloud data set and the environmental image data, and evaluate the performance of the target map model based on the matching results;

[0254] When the performance of the target map model is lower than the preset performance threshold, the accuracy of the target map model is optimized by combining the GPS trajectory and the vehicle communication system V2X.

[0255] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0256] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0257] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0258] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A high-precision map optimization method, characterized in that: The method comprises: collecting raw map-related data from at least one data source; Performing a preprocessing operation on the original map-related data, wherein the preprocessing operation includes cleaning, denoising and time-space alignment; According to the preprocessed map-related data, feature matching is performed between different data sources to obtain a successfully matched feature set; According to the successfully matched feature set, the target map model is trained and obtained; Performing spatial optimization on the target map model to obtain an updated version of the target map model; The updated version of the target map model is pushed to the vehicle system via the cloud.

2. The method according to claim 1, characterized in that Before training and obtaining the target map model according to the successfully matched feature set, the method further includes: Use probabilistic statistical methods and machine learning models to perform data fusion on feature sets; Evaluate the data quality of the fused feature set; When the data quality does not meet the preset mapping standards, the fused feature set is further optimized by using path smoothing and topological structure.

3. The method according to claim 1, characterized in that After the updated version of the target map model is pushed to the vehicle system via the cloud, the following steps are also included: Collect real-time feedback data of the vehicle driving environment; The performance of the target map model is evaluated according to the real-time feedback data of the vehicle driving environment, and the target map model is spatially optimized.

4. The method according to claim 3, characterized in that The real-time feedback data of the vehicle driving environment is collected, including: Identify the vehicle's driving environment and collect real-time feedback data of the vehicle's driving environment; The real-time feedback data is connected to the vehicle system through an API interface and formatted to be compatible with the internal data format of the vehicle system.

5. The method according to claim 4, characterized in that The real-time feedback data of the vehicle driving environment is collected, including: Using laser radar to collect a laser point cloud data set of the driving environment, performing cluster analysis on the laser point cloud data set, and identifying road edges and isolation zones according to the analysis results; A camera device is used to collect environmental image data of the driving environment, and based on the environmental image data, a deep learning algorithm is used to identify traffic signs and lane lines.

6. The method according to claim 5, characterized in that The step of evaluating the performance of the target map model according to the real-time feedback data of the vehicle driving environment and performing spatial optimization on the target map model includes: Performing spatiotemporal alignment of the laser point cloud data set and the environmental image data using the timestamp and location information recorded by the vehicle system; Using a feature matching algorithm, the laser point cloud data set and the environmental image data are matched, and the performance of the target map model is evaluated according to the matching result; When the performance of the target map model is lower than a preset performance threshold, the accuracy of the target map model is optimized by combining the GPS trajectory and the vehicle communication system V2X.

7. A high-precision map optimization device, characterized in that: The device comprises: A data collection module, used for collecting original map-related data from at least one data source; A preprocessing module, used for performing preprocessing operations on the original map-related data, wherein the preprocessing operations include cleaning, denoising and spatiotemporal alignment; A matching module is used to perform feature matching between different data sources based on the preprocessed map-related data to obtain a successfully matched feature set; The model training module is used to train and obtain the target map model according to the successfully matched feature set; A model optimization module, used for performing spatial optimization on the target map model to obtain an updated version of the target map model; The push module is used to push the updated version of the target map model to the vehicle system through the cloud.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images