Generation method and system of high-precision map for simulation

Abstract

The invention relates to the technical field of simulation maps, and discloses a method and system for generating a high-precision map for simulation, and the method comprises the steps: collecting and exporting road information and vehicle information; extracting position information and attitude information in the vehicle information to generate a trajectory; extracting original image data and timestamps in the road information, and extracting vectorization data of pavement marker lines in the original image data; extracting and calculating road boundary data from the image data, and storing the road boundary data according to a time sequence; converting the position coordinates corresponding to the trajectory, the pavement marker line and the road boundary into earth plane coordinates; extracting related data of a trajectory, a pavement marker line and a road boundary, obtaining global coordinates under each frame of data, combining data corresponding to the global coordinates, and constructing a road network relationship; the road network relation is fused with an existing map model, and a high-precision map is generated; the whole implementation process is low in cost, small in calculation amount, capable of being completed through an existing model, high in efficiency and good in implementation effect.

Description

Technical Field

[0001] The present invention relates to the technical field of simulation maps, and in particular to a method and system for generating a high-precision map for simulation. Background Art

[0002] The combination of high-precision maps and virtual simulation technology can accelerate the development of autonomous driving simulation.

[0003] First of all, the simulation test of the autonomous driving system requires the support of high-precision maps. High-precision maps can provide information to the autonomous driving system, allowing the vehicle to drive on actual roads and comply with traffic regulations. They can also serve as a true value test to see whether the autonomous driving system correctly identifies and responds to various situations.

[0004] Secondly, high-precision maps provide rich information about road elements, including road geometry, traffic signs, intersections, lane lines, etc. This information is very important for building test scenarios for autonomous driving systems.

[0005] High-precision maps can provide detailed road information and rich geographical elements such as lane level, curve curvature radius, accurate road shape, lane lines, high-precision coordinates, traffic signs, slope, elevation, roll, etc., which can provide a solid foundation for autonomous driving.

[0006] Currently, high-precision map collection is carried out by high-precision map collection vehicles. The vehicle is equipped with GNSS (Global Navigation Satellite System), IMU (Inertial Measurement Unit), wheel speed meter, laser radar and camera. Among them, GNSS can provide the absolute coordinates of the vehicle, IMU and wheel speed meter can provide the relative position information of the vehicle, and laser radar and camera can provide the three-dimensional environment information around the vehicle. Usually, the cost of a vehicle is more than one million.

[0007] If a 64-line laser radar is used, the two-way lanes are usually fully covered 3 to 5 times during the collection process, preferably 5 times. If a 16-line laser solution is used, more laps are required to achieve better results. Generally, one collection vehicle collects 100 kilometers of effective mileage per day, and the cost can reach 1,000 yuan per kilometer. Because it collects more information, it also requires more back-end processing costs, so the production cycle of the map is long and the efficiency is low. Summary of the invention

[0008] In view of the deficiencies in the prior art, the object of the present invention is to provide a method and system for generating a high-precision map for simulation.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A method for generating a high-precision map for simulation, comprising:

[0011] Collect and export road information and vehicle information;

[0012] Extracting the position information and posture information in the vehicle information to generate a trajectory line;

[0013] Extracting original image data and timestamps from the road information, extracting vectorized data of road surface marking lines from the original image data, and obtaining vertex coordinates of road surface markings corresponding to the road surface marking lines;

[0014] The image data is further processed to extract and calculate the road boundary data and store them in chronological order;

[0015] Converting the position coordinates corresponding to the track line, road marking line, and road boundary into geodetic plane coordinates;

[0016] Extract the trajectory line, road marking line, road boundary related data, obtain the global coordinates under each frame of data, merge the data corresponding to the global coordinates, and construct a road network relationship;

[0017] Integrate road network relationships with existing map models to generate high-precision maps.

[0018] In the present invention, preferably, the generating trajectory line further comprises:

[0019] Acquiring position information and attitude information, wherein the attitude information includes three-axis attitude angle, angular velocity, acceleration, and azimuth, and the position information includes the geographic location coordinates of the vehicle;

[0020] Obtaining a number of positioning information at intervals, the positioning information including longitude, latitude, heading angle, instantaneous speed, and converting the longitude and latitude coordinates into rectangular coordinates of the geodetic plane;

[0021] The point corresponding to the first positioning information is set as the origin, the right side of the vehicle is set as the positive direction of the X axis, and the direction corresponding to the front of the vehicle is set as the positive direction of the Y axis, and the vehicle plane coordinate system is established;

[0022] Other positioning information is converted into the vehicle plane coordinate system according to the corresponding vehicle heading angle to obtain the trajectory line.

[0023] In the present invention, preferably, the road surface marking lines include lane lines, zebra crossings and stop lines; the vectorized data of the road surface marking lines include geometric coordinates of shape points on the road surface edge lines, geometric vectors and attribute information of the road surface edge lines.

[0024] In the present invention, preferably, after extracting the road marking lines, it is also necessary to detect whether the road marking lines constitute a closed figure. If so, the closed figure corresponding to the road marking lines is compared with the image in the national standard road ground marking library. If the similarity is greater than 90%, the recognized closed figure is directly covered with the corresponding standard figure in the national standard road ground marking library, and each vertex of the standard figure is marked clockwise according to the driving direction;

[0025] If the similarity is less than 90%, directly mark the vertices of the closed figure clockwise according to the form direction and mark it as manual processing;

[0026] If the road marking line is not closed, proceed directly to the next step.

[0027] In the present invention, preferably, when the road surface marking line is parallel to the driving trajectory, it is a lane line, the center point of the lane line is marked at a certain interval, and the width of the lane line is recorded at the same time. The lane line is fitted by interpolation method, the noise is removed, and it is stored in chronological order.

[0028] In the present invention, preferably, the extraction and calculation of road boundary data specifically includes converting image data, Gaussian filtering, edge detection, generating Mask, extracting ROI, so as to identify the road boundary, converting the boundary line through the perspective matrix, calculating the road boundary data based on the image pixels, and storing it in chronological order.

[0029] In the present invention, preferably, it also includes identifying objects on both sides of the road and in the middle through a convolutional neural network, wherein the objects include traffic lights, telephone poles, fire hydrants, street lights, and flower beds, and assigning corresponding object labels and recording their size and location data.

[0030] In the present invention, preferably, before obtaining the global coordinates of each frame of data, the trajectory lines, road surface marking lines, and road boundary related data are arranged in chronological order, and linear interpolation is used to align the time series.

[0031] In the present invention, preferably, the Hungarian algorithm is used to match the labels, trajectory lines, road marking lines, and road boundaries in different frames with the existing map model, duplicate objects are filtered out, and the successfully matched objects and the corresponding attribute data of the objects are inserted into the map to complete the fusion, thereby generating a high-precision map.

[0032] A system for generating a high-precision map for simulation includes an acquisition module and a post-processing module. The acquisition module includes a camera component and a navigation system. The camera component is used to collect image data related to road information. The navigation system is used to obtain real-time vehicle information, and the vehicle information includes position information and posture information. The post-processing module includes an image processing group, a trajectory generation group and a fusion group. The image processing group includes a number of image recognition algorithms for processing and identifying input image data to extract road surface marking lines and road boundary related data. The trajectory generation group is used to process vehicle information to generate trajectory lines. The fusion group receives the trajectory lines, surface marking lines, and road boundary related data and processes them to generate a high-precision map.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The method of the present invention collects actual image data and vehicle information, obtains road marking lines and road boundaries through image processing and recognition, obtains trajectory lines through vehicle information calculation, matches and fuses them with existing map models, and generates a high-precision map. The overall implementation process has low cost and small amount of calculation. It can be completed using existing models with high efficiency and good implementation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic diagram of the structure of a method for generating a high-precision map for simulation according to the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0038] A preferred embodiment of the present invention provides a method for generating a high-precision map for simulation, using multiple camera components to collect image data of lane lines, curbs and other road information, using a combined navigation system to collect vehicle information, and then processing the image data to identify lane lines, zebra crossings, stop lines and other road markings, as well as objects on both sides and in the middle of the road including traffic lights, telephone poles, fire hydrants, street lights, flower beds, etc., road boundaries and other data, and obtain trajectory lines by processing vehicle information. Finally, the above data is sorted in chronological order, and the data is merged according to the position coordinates in the data, and matched and fused with the existing map model to generate a high-precision map. The overall implementation process has low cost and small amount of calculation, can be completed using existing models, has high efficiency, and achieves good results.

[0039] The above process will be described in detail below with reference to the accompanying drawings. Figure 1 The method for generating a high-precision map for simulation in this application mainly includes the following steps:

[0040] Collect and export road information and vehicle information;

[0041] Extracting the position information and posture information in the vehicle information to generate a trajectory line;

[0042] Extracting original image data and timestamps from the road information, extracting vectorized data of road surface marking lines from the original image data, and obtaining vertex coordinates of road surface markings corresponding to the road surface marking lines;

[0043] The image data is further processed to extract and calculate the road boundary data and store them in chronological order;

[0044] Converting the position coordinates corresponding to the track line, road marking line, and road boundary into geodetic plane coordinates;

[0045] Extract the trajectory line, road marking line, road boundary related data, obtain the global coordinates under each frame of data, merge the data corresponding to the global coordinates, and construct a road network relationship;

[0046] Integrate road network relationships with existing map models to generate high-precision maps.

[0047] Specifically, the camera component and the integrated navigation system first collect road information and vehicle posture information, and then export the data to the post-processing module for subsequent processing.

[0048] Then, the position information and attitude information in the vehicle information are obtained, which are mainly obtained from the vehicle IMU data and GNSS data collected by the integrated navigation system. The IMU data and GNSS data are used to indicate the vehicle's position information. The vehicle's position information includes the vehicle's position information and attitude information. The IMU data is used to indicate the vehicle's three-axis attitude angle, angular velocity, acceleration, heading angle, and azimuth angle; the GNSS data includes the vehicle's geographic location coordinates. Several positioning information are obtained at intervals. Here, the interval refers to the positioning information of the point taken every time the vehicle moves 0.5m. The positioning information includes longitude, latitude, heading angle, instantaneous speed, and the longitude and latitude coordinates are converted into geodetic plane rectangular coordinates using Gauss Kruger projection or UTM projection. The point corresponding to the first positioning information is set as the origin, and the right side of the vehicle at this time is the positive direction of the X-axis, and the corresponding direction of the front of the vehicle is set as the positive direction of the Y-axis to establish a vehicle plane coordinate system; other positioning information is converted according to the corresponding vehicle heading angle, and converted to the vehicle plane coordinate system to obtain the trajectory line. For example, when the coordinates of the point corresponding to the first positioning information are (x1, y1), and the coordinates of other points are (x2, y2), (x3, y3)..., (xn, yn), the coordinates of any point in the vehicle plane coordinate system are:

[0049]

[0050] Then, the original image data and timestamp in the road information are extracted, and the vectorized data of the road surface marking lines in the image data are identified in turn through superposition edge detection, color filtering, Hough transform, and perspective transform methods. The road surface marking lines include lane lines, zebra crossings, stop lines, etc.; the vectorized data of the road surface marking lines include the geometric coordinates of the shape points on the road surface edge lines, geometric vectors, and attribute information of the road surface edge lines.

[0051] After extracting the road markings, it is also necessary to detect whether the road markings constitute a closed figure. If so, the closed figure corresponding to the obtained road markings is compared with the image in the national standard road ground marking library. If the similarity is greater than 90%, the identified closed figure is directly covered with the corresponding standard figure in the national standard road ground marking library, and the vertices of the standard figure are marked clockwise according to the driving direction; if the similarity is less than 90%, the vertices of the closed figure are directly marked clockwise according to the form direction, and each vertex is recorded as (P1, P2, P3, ..., Pn), and marked as manually processed; if the road markings are not closed, proceed directly to the next step.

[0052] When the road marking line is parallel to the driving trajectory, it is a lane line. The center point of the lane line is marked once every certain interval, such as 10 cm. At the same time, the width B of the lane line is recorded, and the line color data is added. The white line and the yellow line are represented by W and Y respectively. The lane line is recorded as (P1, P2, ..., Pn, B, W / Y). The lane line is fitted by interpolation method, and the noise is removed to make the lane line smooth, and it is stored in chronological order.

[0053] Curbs, green belts in the road, and railings are the boundaries of vehicle travel, which are usually higher than the ground. The image data collected by the binocular camera component is extracted and recorded, and the curb is identified through image conversion, Gaussian filtering, Canny edge detection, Mask mask generation, ROI extraction and other steps. After the camera is calibrated, the boundary line is converted through the perspective matrix, and the road boundary data is calculated according to the image pixels. The above data is stored in chronological order.

[0054] The collected raw image data is processed through a convolutional neural network to identify objects on both sides of the road and in the middle, including traffic lights, telephone poles, fire hydrants, street lights, and flower beds, and corresponding object labels are assigned to record their size and location data.

[0055] After obtaining the relevant data of trajectory lines, road marking lines, and objects on both sides and in the middle of the road, all position coordinates are converted into geodetic plane coordinates and data cleaning is performed. The conversion formula is:

[0056]

[0057] Finally, the trajectory, road markings, and road boundary-related data are extracted and arranged in chronological order. The linear interpolation method is used to align the time series to obtain the global coordinates of each frame of data, and all data are merged through the global coordinates to establish a road network relationship. The data with conflicting information is annotated abnormally, and the actual road conditions are manually confirmed by integrating image information and positioning information.

[0058] The Hungarian algorithm is used to match the labels, trajectory lines, road markings, and road boundaries in different frames with the existing map model, filter out duplicate objects, insert successfully matched objects and their corresponding attribute data into the map, complete the fusion, and generate a high-precision map.

[0059] Some other preferred embodiments of the present invention provide a system for generating a high-precision map for simulation, including an acquisition module and a post-processing module, the acquisition module including a camera component and a navigation system, the camera component is used to collect image data related to road information, the navigation system is used to obtain real-time vehicle information, the vehicle information includes position information and posture information; the post-processing module includes an image processing group, a trajectory generation group and a fusion group, the image processing group includes a number of image recognition algorithms, which are used to process and identify the input image data to extract road surface marking lines and road boundary related data, the trajectory generation group is used to process vehicle information to generate trajectory lines, and the fusion group receives the trajectory lines, surface marking lines, road boundary related data and processes them to generate a high-precision map.

[0060] In some other preferred embodiments of the present invention, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the method described in the above embodiment.

[0061] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0062] The above description is a detailed description of the preferred feasible embodiments of the present invention, but the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modified changes completed under the technical spirit suggested by the present invention should fall within the patent scope covered by the present invention.

Claims

1. A method for generating high-precision maps for simulation, It is characterized in that include: Collect and export road information and vehicle information; Extracting the position information and posture information in the vehicle information to generate a trajectory line; Extracting original image data and timestamps from the road information, extracting vectorized data of road surface marking lines from the original image data, and obtaining vertex coordinates of road surface markings corresponding to the road surface marking lines; The image data is further processed to extract and calculate the road boundary data and store them in chronological order; Converting the position coordinates corresponding to the track line, road marking line, and road boundary into geodetic plane coordinates; Extract the trajectory line, road marking line, road boundary related data, obtain the global coordinates under each frame of data, merge the data corresponding to the global coordinates, and construct a road network relationship; Integrate road network relationships with existing map models to generate high-precision maps.

2. A method for generating a high-precision map for simulation according to claim 1, It is characterized in that The generating trajectory line specifically includes: Acquiring position information and attitude information, wherein the attitude information includes three-axis attitude angle, angular velocity, acceleration, and azimuth, and the position information includes the geographic location coordinates of the vehicle; Obtaining a number of positioning information at intervals, the positioning information including longitude, latitude, heading angle, instantaneous speed, and converting the longitude and latitude coordinates into rectangular coordinates of the geodetic plane; The point corresponding to the first positioning information is set as the origin, the right side of the vehicle is set as the positive direction of the X axis, and the direction corresponding to the front of the vehicle is set as the positive direction of the Y axis, and the vehicle plane coordinate system is established; Other positioning information is converted into the vehicle plane coordinate system according to the corresponding vehicle heading angle to obtain the trajectory line.

3. A method for generating a high-precision map for simulation according to claim 2, It is characterized in that The road surface marking lines include lane lines, zebra crossings and stop lines; the vectorized data of the road surface marking lines include geometric coordinates of shape points on the road surface edge lines, geometric vectors and attribute information of the road surface edge lines.

4. A method for generating a high-precision map for simulation according to claim 3, It is characterized in that After extracting the road markings, it is also necessary to detect whether the road markings form a closed figure. If so, the closed figure corresponding to the road markings is compared with the image in the national standard road ground marking library. If the similarity is greater than 90%, the recognized closed figure is directly covered with the corresponding standard figure in the national standard road ground marking library, and each vertex of the standard figure is marked clockwise according to the driving direction; If the similarity is less than 90%, directly mark the vertices of the closed figure clockwise according to the form direction and mark it as manual processing; If the road marking line is not closed, proceed directly to the next step.

5. A method for generating a high-precision map for simulation according to claim 4, It is characterized in that When the road marking line is parallel to the driving trajectory, it is a lane line. The center point of the lane line is marked at a certain interval, and the width of the lane line is recorded at the same time. The lane line is fitted by interpolation method, the noise is removed, and it is stored in chronological order.

6. A method for generating a high-precision map for simulation according to claim 1, It is characterized in that The extraction and calculation of road boundary data specifically includes converting image data, Gaussian filtering, edge detection, generating Mask, extracting ROI, thereby identifying road boundaries, converting boundary lines through a perspective matrix, calculating road boundary data based on image pixels, and storing them in chronological order.

7. The method for generating a high-precision map for simulation according to claim 1, It is characterized in that It also includes using a convolutional neural network to identify objects on both sides of the road and in the middle, including traffic lights, telephone poles, fire hydrants, street lights, and flower beds, and assigning corresponding object labels and recording their size and location data.

8. The method for generating a high-precision map for simulation according to claim 1, It is characterized in that Before obtaining the global coordinates of each frame of data, the trajectory lines, road marking lines, and road boundary related data are arranged in chronological order, and linear interpolation is used to align the time series.

9. The method for generating a high-precision map for simulation according to claim 1, It is characterized in that The Hungarian algorithm is used to match the labels, trajectory lines, road markings, and road boundaries in different frames with the existing map model, filter out duplicate objects, insert successfully matched objects and their corresponding attribute data into the map, complete the fusion, and generate a high-precision map.

10. A system for generating a high-precision map for simulation, used in the method for generating a high-precision map for simulation as claimed in any one of claims 1 to 9. It is characterized in that It includes an acquisition module and a post-processing module. The acquisition module includes a camera component and a navigation system. The camera component is used to collect image data related to road information. The navigation system is used to obtain real-time vehicle information, and the vehicle information includes position information and posture information. The post-processing module includes an image processing group, a trajectory generation group and a fusion group. The image processing group includes several image recognition algorithms for processing and identifying input image data to extract road surface marking lines and road boundary related data. The trajectory generation group is used to process vehicle information to generate trajectory lines. The fusion group receives the trajectory lines, surface marking lines, and road boundary related data and processes them to generate high-precision maps.

Images