Method for constructing, compiling and reading small-volume high-precision positioning layer
Technical Field
The invention relates to the technical field of high-precision maps applied to automatic driving, in particular to a method for constructing, compiling and reading a small-volume high-precision positioning map layer.
Background
Autopilot, broadly refers to a technique that assists or replaces human driving of an automobile. With the development of the technology, the travel of people is more convenient, the influence of human factors of manual driving is reduced, and the driving safety can be further improved to a certain degree. Among the techniques of autopilot, high-precision positioning is important because it directly affects the inputs of other autopilot modules. Accurate positioning is a prerequisite for performing other autonomous driving functions such as sensing and decision control. The automatic driving positioning depends on the combination of vision or laser radar and a high-definition map, namely, the real-time perceived road elements are compared with prior information in the map to perform transverse and longitudinal deviation correction. The map information is required to be stored in a reasonable storage format, and the map information contains enough map content, and meanwhile, the requirements of storage space, retrieval speed, adaptation with a positioning algorithm and the like are also required to be considered.
The existing high-precision map for positioning mainly adopts two ways to format: a topological vector map, the map is similar to the format of the traditional navigation map, the volume is small, but there is not enough information for positioning, can't form the positioning mode through matching of the characteristic point; the other is a panoramic point cloud or a feature point map layer, and although the map directly describes environment information, a high-precision positioning algorithm through feature point matching can be realized, the map has a large volume, occupies a large storage space, and is limited in retrieval speed.
Disclosure of Invention
An object of the present invention is to solve at least the above problems and to provide at least the advantages described later.
The invention also aims to provide a construction method of a small-volume high-precision positioning layer, and a map format with clear specification and taking account of map data volume and reading efficiency is realized.
In order to achieve the above objects and other objects, the present invention adopts the following technical solutions:
a construction method of a small-volume high-precision positioning layer comprises the following steps:
s1, collecting the three-dimensional marker information and lane line information contained in the constructed positioning map area by using a laser radar and a camera respectively, and describing in a characteristic point form;
s2, compressing the content described in the S1 in the form of the characteristic points to form a binary data packet;
s3, projecting the binary data packet to a vector map;
and S4, comprehensively compiling the three-dimensional marker information and the lane line information which are described in the form of the feature points and the vector map.
Preferably, in the method for constructing the small-volume high-precision positioning layer, the three-dimensional marker information is acquired by a laser radar and then described in the form of point cloud feature points.
Preferably, in the method for constructing a small-volume high-precision positioning layer, the three-dimensional marker includes one or more of the following: signpost, median, curb, portal frame and street lamp.
Preferably, in the method for constructing the small-volume high-precision positioning layer, the acquiring of the lane line information by using a laser radar and a camera respectively comprises:
s1-1, acquiring three-dimensional position information of head and tail points of each virtual line segment of the lane line by using a camera;
and S1-2, collecting point cloud data of the virtual line segment of the lane line by using a laser radar.
Preferably, in the method for constructing a small-volume high-precision positioning layer, a compression algorithm meeting the requirement of the positioning algorithm is adopted in S2 to compress the content described in the form of the feature points in S1.
A compiling method of a small-volume high-precision positioning layer comprises the following steps:
step 1, dividing a small-volume high-precision positioning layer constructed by the method of any one of claims 1-5 into at least 16 levels of tile structures, and numbering the tiles;
step 2, identifying feature point information of the positioning layer elements contained in each tile;
and 3, compiling each positioning layer element into corresponding positioning layer data according to the tiles.
Preferably, in the method for compiling the small-volume high-precision positioning layer, the feature point of each positioning layer element is associated with at least one tile.
A reading method of a small-volume high-precision positioning layer comprises the following steps:
a, calculating tile numbers of the periphery of the current position, and acquiring attribute information of the positioning layer element from a database I by using the tile numbers;
b, acquiring feature point data of the positioning layer element from a database II by utilizing the attribute information of the positioning layer element;
the database I is a database which is created by taking the serial number of the tile as an index and is used for searching attribute information of the element of the positioning layer according to the serial number;
the database II is a database which is created by taking the attribute information of the positioning layer elements as an index and used for storing the binary data packets obtained by the method in the claim 1.
The invention at least comprises the following beneficial effects:
the construction method of the small-volume high-precision positioning layer comprises the steps of firstly utilizing a laser radar and a camera to respectively obtain three-dimensional marker information and lane line information, then compressing the obtained characteristic information, projecting the compressed characteristic information onto a vector map, finally comprehensively compiling the positioning layer information and the vector map to generate a positioning layer, respectively obtaining the positioning layer through the laser radar and the camera, enabling the finally generated positioning layer to contain richer map contents and have higher precision, simultaneously obtaining the three-dimensional marker information and the lane information, namely only operating map elements such as lane lines, guideboards and the like belonging to roads and matching the vector map for use, avoiding the defect that the traditional positioning map needs to take the surrounding environment such as tall building trees and the like as collected and sampled and stored, saving a large amount of storage space, the generated positioning layer is smaller in size, so that a high-definition map format which is small in size and refined to the lane level is realized.
The positioning layer is divided into the tile structures by the compiling method, so that the positioning layer is clear in specification.
By setting the database and the compiling method for dividing the positioning layer into the tile structure, the topological property of the vector map and the point cloud map of map elements are comprehensively compiled, the map volume is greatly reduced, and the map calling speed is obviously improved.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention.
Detailed Description
The present invention is described in detail below to enable one skilled in the art to practice the invention in light of the description.
A construction method of a small-volume high-precision positioning layer comprises the following steps:
s1, collecting the three-dimensional marker information and lane line information contained in the constructed positioning map area by using a laser radar and a camera respectively, and describing in a characteristic point form;
s2, compressing the content described in the S1 in the form of the characteristic points to form a binary data packet;
s3, projecting the binary data packet to a vector map;
and S4, comprehensively compiling the three-dimensional marker information and the lane line information which are described in the form of the feature points and the vector map.
In the scheme, the three-dimensional marker information and the lane line information are respectively obtained by utilizing the laser radar and the camera, the obtained characteristic information is compressed and projected onto the vector map, and finally the positioning map layer information and the vector map are comprehensively compiled to generate the positioning map layer which is respectively obtained by the laser radar and the camera, so that the finally generated positioning map layer contains richer map contents and has higher precision, and the three-dimensional marker information and the lane information are simultaneously obtained, namely, the operation is only carried out on map elements such as lane lines, guideboards and the like which belong to roads, and the positioning map layer is matched with the vector map for use, thereby avoiding the defect that the traditional positioning map needs to take the surrounding environments such as tall trees, and the like as positioning data for collecting, sampling and storing, saving a large amount of storage space and ensuring that the generated positioning map layer is smaller in size, therefore, a high-definition map format which is small in size and refined to the lane level is achieved.
The positioning layer is divided into the tile structures by the compiling method, so that the positioning layer is clear in specification.
By setting the database and the compiling method for dividing the positioning layer into the tile structure, the topological property of the vector map and the point cloud map of map elements are comprehensively compiled, the map volume is greatly reduced, and the map calling speed is obviously improved.
In a preferred scheme, the three-dimensional marker information is acquired by using a laser radar and then described in the form of point cloud characteristic points.
In a preferred embodiment, the three-dimensional identifier comprises one or more of: signpost, median, curb, portal frame and street lamp.
In the above solution, the three-dimensional marker includes one or more of a guideboard, a median, a kerb, a portal frame and a street lamp, and is not limited to the above-mentioned ones.
In a preferred embodiment, the collecting the lane line information by using a laser radar and a camera respectively comprises:
s1-1, acquiring three-dimensional position information of head and tail points of each virtual line segment of the lane line by using a camera;
and S1-2, collecting point cloud data of the virtual line segment of the lane line by using a laser radar.
In a preferred embodiment, the content described in S1 in the form of feature points is compressed by using a coincidence localization algorithm in S2.
A compiling method of a small-volume high-precision positioning layer comprises the following steps:
step 1, dividing a small-volume high-precision positioning layer constructed by the method of any one of claims 1-5 into at least 16 levels of tile structures, and numbering the tiles;
step 2, identifying feature point information of the positioning layer elements contained in each tile;
and 3, compiling each positioning layer element into corresponding positioning layer data according to the tiles.
In the scheme, the obtained content described in the form of the feature points is convenient to compress and convert into binary database data by compiling the positioning layer; meanwhile, the purpose of combining the positioning layer and the vector layer is achieved.
In a preferred embodiment, each feature point of the positioning layer element is associated with at least one tile.
In the above scheme, the elements of one positioning layer are allowed to associate multiple tiles in consideration of the possible cross-tile situation.
A reading method of a small-volume high-precision positioning layer comprises the following steps:
a, calculating tile numbers of the periphery of the current position, and acquiring attribute information of the positioning layer element from a database I by using the tile numbers;
b, acquiring feature point data of the positioning layer element from a database II by utilizing the attribute information of the positioning layer element;
the database I is a database which is created by taking the serial number of the tile as an index and is used for searching attribute information of the element of the positioning layer according to the serial number;
the database II is a database which is created by taking the attribute information of the positioning layer elements as an index and used for storing the binary data packets obtained by the method in the claim 1.
Examples
The method for constructing, compiling and reading the small-volume high-precision positioning layer is adopted to reconstruct the map grid and compile the positioning layer, and the specific implementation mode is as follows:
1. transformation of map grid
A. Information such as three-dimensional markers (guideboards, isolation strips and the like) is described in the form of characteristic points of laser radar point cloud.
B. Information such as three-dimensional markers (guideboards, isolation belts and the like) is described in the form of characteristic points collected by a camera.
C. And describing the three-dimensional position information of the ending point of each virtual line segment of the lane line, which is acquired by the camera.
D. Point cloud data of a virtual line segment of a lane line acquired by a laser radar is described.
E. And compressing the map content described in the ABCD according with a positioning algorithm to form a binary data packet.
F. And projecting the compressed feature point content on a vector map to refine to a lane level.
G. And comprehensively compiling the positioning layer information and the traditional vector map.
2. Compilation of positioning layers
A. The map data is divided into tile structures of more than 16 levels.
B. And judging the tile where the feature point of each positioning layer element is located.
C. Elements of a positioning layer are allowed to associate with multiple tiles, taking into account possible cross-tile situations.
D. Compiling the tile number covered by each positioning layer element into attribute information of the element.
E. And creating a database by taking the tile numbers as indexes for finding the positioning layer element numbers according to the tile numbers.
F. And creating a database for the index according to the serial number of the element of the positioning layer, and storing the compressed feature point information.
G. When the positioning layer elements are extracted, a peripheral tile number list is quickly calculated according to the current position to search a database E, and the numbers of the positioning layer elements are obtained according to the list.
H. And searching a database F according to the serial number of the element of the positioning layer, and realizing the extraction of the feature point data of the element of the positioning layer.
While embodiments of the invention have been disclosed above, it is not limited to the applications listed in the description and the embodiments, which are fully applicable in all kinds of fields of application of the invention, and further modifications may readily be effected by those skilled in the art, so that the invention is not limited to the specific details without departing from the general concept defined by the claims and the scope of equivalents.