阅读说明：本技术 一种局部高精度地图的快速生成和制作方法 (Rapid generation and manufacturing method of local high-precision map ) 是由 彭莎莎 汪红刚 程华山 祝鹏 于 2021-08-30 设计创作，主要内容包括：本发明涉及计算机仿真技术领域,且公开了一种局部高精度地图的快速生成和制作方法,包括以下步骤,S1,通过RGB相机,基于相机数据,将采集图像分割为两个区域：道路区域和非道路区域,S2,结合毫米波雷达、激光雷达融合来分割出道路的点云,生成车道点云；S3,基于这些参数实现部署到周围的传感器组合,包括毫米波雷达、激光雷达、RGB相机等,采集充足的原始数据,进行结构化分析,生成车道高清地图。该一种局部高精度地图的快速生成和制作方法,能够从毫米波雷达、激光雷达、RGB相机等原始传感器数据中通过计算机自动生产高精地图,为自动驾驶车辆的部署做出了有意义的贡献。(The invention relates to the technical field of computer simulation, and discloses a method for quickly generating and manufacturing a local high-precision map, which comprises the following steps of S1, dividing an acquired image into two areas based on camera data through an RGB camera: dividing point clouds of roads by combining millimeter wave radar and laser radar to generate lane point clouds in S2; and S3, based on the parameters, the sensor combination deployed to the periphery is realized, including millimeter wave radar, laser radar, RGB cameras and the like, sufficient original data are collected, structured analysis is carried out, and a lane high-definition map is generated. The method for quickly generating and manufacturing the local high-precision map can automatically produce the high-precision map from original sensor data such as millimeter wave radar, laser radar, RGB (red, green and blue) cameras and the like through a computer, and makes a meaningful contribution to the deployment of automatic driving vehicles.)

1. A method for quickly generating and manufacturing a local high-precision map is characterized by comprising the following steps: comprises the following steps of (a) carrying out,

for a detected road, dividing an acquired image into two areas through an RGB camera based on camera data: road and non-road regions, which will produce a binary image;

secondly, dividing point clouds of the road by combining millimeter wave radar and laser radar fusion, realizing accurate measurement of vehicles, pedestrians, roads, warning signs and the like in the current area, forming accurate road condition structured information of the current area, and generating lane point clouds;

and thirdly, realizing sensor combination deployed to the periphery based on the parameters, including a millimeter wave radar, a laser radar, an RGB camera and the like, acquiring sufficient original data, performing structural analysis, forming comprehensive description of people, objects and scenes of the local space, forming continuous distribution of objects and events of the local space in a space-time coordinate system, and generating a lane high-definition map.

2. The method for rapidly generating and manufacturing the local high-precision map according to the first step of claim 1, wherein the method comprises the following steps: for detecting roads, a full convolution network is used, applied to the RGB camera data, splitting the image into two regions: and a road area and a non-road area, which generate a binary image, and then the point cloud of the road is segmented by combining the millimeter wave radar and the laser radar.

3. The method for rapidly generating and manufacturing the local high-precision map according to claim 2, wherein the method comprises the following steps: for lane detection, LaneNET is used, all lanes visible to RGB can be detected by selecting the network, not only the lane of the current vehicle, and the mask image is combined with millimeter wave radar and laser radar calibration to generate lane point cloud.

4. The method for rapidly generating and manufacturing the local high-precision map according to the step two of claim 1, wherein the step two comprises the following steps: the farther the millimeter wave radar and the laser radar are away from the automobile, the lower the calibration precision is, so that the 'camera field point cloud' is firstly cut to a position at a certain distance L from the Fl frame origin, then the lane mask is projected on the 'camera field point cloud', and finally, the points belonging to the lane are extracted by using a color segmentation method to form the lane point cloud.

5. The method for rapidly generating and manufacturing the local high-precision map according to claim 1, wherein the method comprises the following steps: the generated lane point cloud is noisy, and a series of clustering and smoothing steps are established after the lane point cloud is generated, and the steps are applied to the lane point cloud to generate a series of path points, and the path points can be used by the autonomous automobile to know the position of a lane in the space.

Technical Field

The invention relates to the technical field of computer simulation, in particular to a method for quickly generating and manufacturing a local high-precision map.

Background

With the rapid development of computer technology, the application of maps is becoming widespread. For example, high-precision maps play a significant role in unmanned vehicle automatic driving systems. In particular, in unmanned vehicle autopilot systems, both perception, path planning, and positioning systems rely on high-precision maps to varying degrees.

The related art mainly relies on the high-precision global positioning result provided by the GNSS/SINS integrated navigation system to generate the map.

The prior art has the following defects and shortcomings:

in a world where autonomous vehicles are becoming more prevalent, it is crucial to create a sufficient infrastructure for this new technology, which includes accurate and efficient construction of high-accuracy maps, and today, the process of creating high-accuracy maps requires a lot of manual effort, which is both time-consuming and error-prone.

Disclosure of Invention

Aiming at the defects of the prior art, the invention provides a method for quickly generating and manufacturing a local high-precision map, which can solve the problems that the existing process for manufacturing the high-precision map needs a large amount of labor investment, the mode needs time and is easy to make mistakes; according to the invention, through parameter description of the local space high-precision map, sensor combinations (millimeter wave radar, laser radar, RGB camera and the like) deployed to the periphery are realized based on the parameters, sufficient original data are collected and structurally analyzed to form comprehensive description of people, objects and scenes of the local space, and continuous distribution of objects and events in a space-time coordinate system of the local space is formed, so that the high-precision map can be automatically produced from the original sensor data through a computer, and meaningful contribution is made to the deployment of automatic driving vehicles.

In order to achieve the purpose of the method for quickly generating and manufacturing the local high-precision map, the invention provides the following technical scheme: a method for quickly generating and manufacturing a local high-precision map comprises the following steps,

for a detected road, dividing an acquired image into two areas through an RGB camera based on camera data: road and non-road regions, which will produce a binary image;

secondly, dividing point clouds of the road by combining millimeter wave radar and laser radar fusion, realizing accurate measurement of vehicles, pedestrians, roads, warning signs and the like in the current area, forming accurate road condition structured information of the current area, and generating lane point clouds;

and thirdly, realizing sensor combination deployed to the periphery based on the parameters, including a millimeter wave radar, a laser radar, an RGB camera and the like, acquiring sufficient original data, performing structural analysis, forming comprehensive description of people, objects and scenes of the local space, forming continuous distribution of objects and events of the local space in a space-time coordinate system, and generating a lane high-definition map.

Preferably, for detecting roads, a full convolution network is used, which is applied to the RGB camera data, dividing the image into two regions: and a road area and a non-road area, which generate a binary image, and then the point cloud of the road is segmented by combining the millimeter wave radar and the laser radar.

Preferably, for lane detection, LaneNET is used, all lanes visible to RGB can be detected by the network, not only the lane of the current vehicle, and the mask image is combined with millimeter wave radar and laser radar calibration to generate a lane point cloud.

Preferably, the farther the millimeter wave radar and the laser radar are away from the automobile, the lower the calibration precision is, so that the 'camera field point cloud' is firstly cut to a position at a certain distance L from the Fl frame origin, then a lane mask is projected on the 'camera field point cloud', and finally, points belonging to a lane are extracted by using a color segmentation method to form the lane point cloud.

Preferably, the generated lane point cloud is noisy, and a series of clustering and smoothing steps are established after the lane point cloud is generated, and the steps are applied to the lane point cloud to generate a series of path points, and the path points can be used by the autonomous vehicle to know the position of the lane in the space.

Compared with the prior art, the invention provides a method for quickly generating and manufacturing a local high-precision map, which has the following beneficial effects:

the method for quickly generating and manufacturing the local high-precision map comprises the steps of describing parameters of the local space high-precision map, realizing that sensor groups deployed to the periphery comprise a millimeter wave radar, a laser radar, an RGB camera and the like based on the parameters, collecting sufficient original data through the millimeter wave radar, the laser radar and the RGB camera, carrying out structural analysis, forming comprehensive description of people, objects and scenes of the local space, forming continuous distribution of objects and events of the local space in a space-time coordinate system, automatically producing the high-precision map through a computer from the original sensor data of the millimeter wave radar, the laser radar, the RGB camera and the like, and making meaningful contribution to the deployment of automatic driving vehicles.

Drawings

FIG. 1 is a schematic view of the overall flow structure of the present invention;

fig. 2 is a schematic view of the overall frame structure of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Referring to fig. 1-2, a method for rapidly generating and manufacturing a local high-precision map includes the following steps, firstly, for a detected road, dividing an acquired image into two regions based on camera data by an RGB camera: the method comprises the steps of generating a binary image in a road area and a non-road area, then combining a millimeter wave radar and a laser radar to divide point cloud of a road, realizing accurate measurement of vehicles, pedestrians, roads, warning signs and the like in the current area, forming accurate road condition structural information of the current area, generating lane point cloud, finally realizing sensor combination deployed to the periphery based on the parameters, including the millimeter wave radar, the laser radar, an RGB camera and the like, collecting sufficient original data, carrying out structural analysis, forming comprehensive description of people, objects and scenes of the local space, forming continuous distribution of objects and events of the local space in a space-time coordinate system, and generating a lane high-definition map.

In summary, for detecting roads, a full convolution network is used, which is applied to the RGB camera data, splitting the image into two regions: a road area and a non-road area, which will generate a binary image, then a point cloud of the road will be divided by combining the millimeter wave radar and the laser radar,

firstly, cutting point cloud by using RGB camera parameters, thus only operating points in the field of view of the camera, and then projecting a binary image onto the point cloud by using external parameters between a millimeter wave radar and a laser radar;

for lane detection, LaneNET is used, the network is selected to be capable of detecting all lanes visible to an RGB camera, not just the lane of a current vehicle, and a mask image is combined with millimeter wave radar and laser radar calibration to generate lane point cloud;

all lanes visible from the RGB camera, not just the lane of the current vehicle, can be detected using LaneNET;

the farther the millimeter wave radar and the laser radar are from the automobile, the lower the calibration precision is, so that firstly, the 'camera field point cloud' is cut to a position at a certain distance L from the Fl frame origin, then a lane mask is projected on the 'camera field point cloud', finally, points belonging to a lane are extracted by using a color segmentation method to form a lane point cloud,

sufficient original data are collected through a millimeter wave radar, a laser radar and an RGB camera, structured analysis is carried out, comprehensive description of people, objects and scenes of the local space is formed, and continuous distribution of objects and events of the local space in a space-time coordinate system is formed;

the generated lane point cloud is noisy, and a series of clustering and smoothing steps are established after the lane point cloud is generated, the steps are applied to the lane point cloud, a series of path points are generated, the path points can be used for an autonomous automobile to know the position of a lane in the space,

noise is reduced by building a series of clustering and smoothing steps.

The working use process and the installation method of the invention are that when the rapid generation and manufacturing method of the local high-precision map is used, the high-precision map of the local space is described by parameters of the local high-precision map, sensor groups deployed to the periphery comprise a millimeter wave radar, a laser radar, an RGB camera and the like are realized based on the parameters, sufficient original data are collected by the millimeter wave radar, the laser radar and the RGB camera, the structural analysis is carried out, the comprehensive description of people, objects and scenes of the local space is formed, the continuous distribution of objects and events in a space-time coordinate system of the local space is formed, the high-precision map can be automatically produced from the original sensor data of the millimeter wave radar, the laser radar, the RGB camera and the like through a computer, and the significant contribution is made to the deployment of automatic driving vehicles.

It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.

Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.