OSM-class map-based SLAM map global alignment system

Abstract

An SLAM map global alignment system based on an OSM map comprises a priori OSM map processing module, an SLAM map data processing module, a global map coarse alignment module and a map optimization-based fine alignment module, a topological map is constructed as a transfer structure of map alignment by using a common topological structure in a priori map as a key point, and the overall alignment accuracy of the map alignment is improved. The effects of aligning the perception map with the road network map and endowing the SLAM map with global coordinates are achieved, global coordinate calibration, namely geographic longitude and latitude coordinates, of the SLAM map can be completed with relatively high precision, semantic information association is achieved after the SLAM map is aligned with the OSM, and the method has relatively good application extensibility for future map application.

Description

Technical Field

[0001] The present invention relates to a technology in the field of intelligent driving, specifically a SLAM (Simultaneous Localization and Mapping) map global alignment system based on an open source map (OpenStreet Map) type map. Background Art

[0002] The maps built by existing positioning and mapping technologies usually only have local information in the vehicle coordinate system, and lack global information aligned with the world coordinate system, which is far from enough for actual driving scenarios. In completely unfamiliar scenarios, without the assistance of mapping or high-definition maps, SLAM technology is basically unable to obtain global coordinate position coordinates, and high-definition maps are Figure 1 Generally, it requires a lot of investment and precise mapping in the early stage, which undoubtedly greatly increases the cost of SLAM map application. Therefore, an automatic map alignment method that can give SLAM map global coordinates at low cost is indispensable for the industrial application of SLAM map. Summary of the invention

[0003] In view of the problem that the map constructed by the existing SLAM technology only has local information in the car coordinate system and lacks global information aligned with the world coordinate system, resulting in the inability to obtain global coordinate position coordinates in unfamiliar scenes or without the assistance of mapping or high-definition maps, the present invention proposes a SLAM map global alignment system based on OSM-type maps, which uses the common topological structure in the prior map as the key point to construct a topological map as a transit structure for map alignment, so as to align the perception map with the road network map and give the SLAM map global coordinates. The global coordinate calibration of the SLAM map, i.e. the longitude and latitude coordinates in the geographical sense, can be completed with relatively high accuracy, and the semantic information after the SLAM map is aligned with the OSM is associated, which has good application scalability for future map applications.

[0004] The present invention is achieved through the following technical solutions:

[0005] The invention relates to a SLAM map global alignment system based on an OSM map, comprising: an a priori OSM map processing module, a SLAM map data processing module, a global map coarse alignment module and a fine alignment module based on graph optimization, wherein: the a priori OSM map processing module extracts a priori topological map with semantic nodes meaningful to driving from original map data; the SLAM map data processing module obtains SLAM map data through a SLAM algorithm, performs sparse processing and key point extraction, and then generates a SLAM topological information map with key points; the global data coarse alignment module performs coarse alignment processing on the a priori topological map and the SLAM topological information map, and filters out topological map node pairs that meet graph optimization fine alignment conditions; the graph optimization fine alignment module constructs a graph optimization system according to map node data to be aligned, optimizes and adjusts the initially constructed SLAM topological information map, obtains an optimized SLAM topological information map, and obtains a SLAM map containing global coordinates after inverse processing by the SLAM map data processing module.

[0006] The prior topological map G map = {V map ,E map}, V map and E map They represent nodes and edges in the topological map respectively.

[0007] The SLAM topology information map G SLAM = {V SLAM ,E SLAM}, V SLAM and E SLAM They represent nodes and edges in the SLAM topology map respectively.

[0008] The original map data refers to: a priori OSM-type road network map, that is, map data in a format similar to OpenStreetMap, including: recording the latitude and longitude coordinates of a certain point in the map and semantic information Nodes, marked as Contains road or contour information Ways composed of Nodes, marked as And store possible related information Relations between each information.

[0009] The extraction refers to: the prior OSM map processing module analyzes the latitude and longitude coordinates and semantic information of the map midpoint in the original map data and road or contour (such as walls, etc.) information According to TAGS N seg and TAGSW seg Extract vehicle accessible roads After that, extract the As V map , retaining semantic information and building on the relationship information Relations Topological relationship between each other map , forming a topological semantic map G map = {V map ,E map}.

[0010] The SLAM map data refers to: the SLAM positioning trajectory established by the existing SLAM algorithm (ORB-SLAM of camera sensor, LOAM of radar sensor, etc.) and the surrounding environment information, mainly in the form of point cloud, and the map contains the local coordinates of the map points.

[0011] The sparse processing refers to: extracting fewer key points from the places without curvature changes in the obtained SLAM map data to achieve the effect of sparse processing and reduce the amount of data.

[0012] The key point extraction is to select a fixed time length T as a sliding window, take the direction of the first positioning point X0 within T time as the positive direction, fit the positioning points within T time as a curve, and mark the extreme point of the curve as a key point V SLAM , which is the inflection point in the true sense, and sparsely select data according to the curvature to reduce the computational burden of later optimization.

[0013] The rough alignment process includes: topological information alignment and semantic information alignment.

[0014] Topology information alignment means: comparing the V map and V SLAM The degree of the node and the structure of a certain level of subtrees expanded with itself as the root node are used as the first level of alignment judgment based on the difference in degree and the similarity of the subtree structure (detected using the WL-Test method);

[0015] Semantic information alignment means: map In the case of containing semantic information, the matching V SLAM Determine whether it is possible to have the semantic information, such as V with intersection semantic labels map More likely to match a V with a larger curvature SLAM .

[0016] The map node data to be aligned refers to: topological map node pairs (V match map ,V match SLAM), which is the result of rough alignment

[0017] The topological map node pair (V match map ,V match SLAM ), V match map and V match SLAM The number of elements in the two node sets is the same and they have a one-to-one matching relationship.

[0018] The optimization adjustment mentioned above refers to: optimizing the system G according to the graph match = {V match ,E match}, V match =V SLAM ∪V match map , keep V match map The node attributes in the set are fixed, which is used as one of the optimization benchmarks; match =Relation(V match map ,V match SLAM )∪E SLAM , the edges in the SLAM map data obtained by the graph optimization fine alignment module and the SLAM topology information map obtained by the SLAM map data processing module are again used as the input of the graph optimization fine alignment module, and iterative optimization is performed until the maximum number of iterations I is reached max Or the optimization error is less than the error threshold Error th So far, the optimized SLAM topology information map is obtained.

[0019] The optimization error is: to meet E match The value of the loss function as the optimization target. When there is no error, it means that there is a set of solutions that can perfectly satisfy E match , but it is generally not possible. The optimization error is the value of the loss function at the current number of iterations.

[0020] The inverse processing refers to: restoring the optimized SLAM topology information map to the original SLAM map format, including the coordinate position changes in the SLAM trajectory and environmental information, and the addition of global coordinate information. Technical Effects

[0021] The present invention uses the structure of a graph to match two heterogeneous map data, OSM-like road network map and SLAM perception map, and optimizes the SLAM perception map in combination with a graph optimization algorithm and assigns global position coordinates to the SLAM mapping result. The present invention uses lightweight original OSM-like road network map data to achieve the effect of assigning global position coordinates to the SLAM mapping result. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the system of the present invention;

[0023] Figure 2 It is a schematic diagram of the effect of the embodiment;

[0024] In the figure: (a) is the visualization of the prior topological map; (b) is the visualization of the SLAM topological map, where the yellow bold nodes are key nodes; (c) is the visualization result after alignment. DETAILED DESCRIPTION

[0025] like Figure 1 As shown, this embodiment relates to a SLAM map global alignment system based on OSM-like maps, including: a priori OSM-like map processing module, a SLAM map data processing module, a global map coarse alignment module and a fine alignment module based on graph optimization.

[0026] The prior OSM map processing module includes: a data splitting unit and a map generating unit, wherein: the data splitting unit generates a map based on the public map data, the map data obtained in the previous experimental process or the prior information D map The data is split according to the category to obtain the longitude and latitude coordinates in the map as well as the semantic information Nodes, road or contour information Ways and related information Relations. The graph generation unit generates the map based on Nodes, Ways and Relations. and Analyze and process the information of the category according to TAGS N seg and TAGS W seg Extract vehicle accessible roads Then the road Extract it as graph node V map , retaining its semantic information and building V according to Relations map The topological relationship between them, as the edge E map Get the prior topological map G map = {V map ,E map}result.

[0027] The SLAM map data processing module includes: a key point extraction unit and a graph generation unit, wherein: the key point extraction unit selects a fixed time length T as a sliding window according to the SLAM map information, takes the direction of the first positioning point X0 within T time as the positive direction, fits the positioning points within T time as a curve, and marks the extreme point of the curve as a key point V SLAM , which is the inflection point in the true sense, and sparsely selects data according to the curvature size to reduce the computational burden of later optimization. The graph generation unit selects the key point V according to the key point V. SLAM And the key point V in the SLAM map SLAM Related semantic information and local location information are used to construct the initial SLAM topology map G SLAM = {V SLAM ,E SLAM}.

[0028] The input of the global map coarse alignment module is the output of the first two modules {G map ,G SLAM}, this module will perform coarse alignment on the input data and filter out the topological map node pairs that meet the conditions for fine alignment of graph optimization (V match map ,V match SLAM ), V match map and V match SLAM The number of elements in the two node sets is the same and they have a one-to-one matching relationship. The alignment logic includes two parts: topological information and semantic information. Topological information alignment is to compare the V map and V SLAM The degree of G and the structure of a certain level of subtrees expanded with itself as the root node are used as the first level of alignment judgment based on the difference in degree and the similarity of the subtree structure (detected using the WL-Test method); semantic information alignment is based on G map In the case of containing semantic information, the matching V SLAM Determine whether it is possible to have the semantic information, such as V with intersection semantic labels map More likely to match a V with a larger curvature SLAM .

[0029] After rough alignment (V match map ,V match SLAM ) The node pairs have given the SLAM map global coordinates to a certain extent, but the alignment results need to be further optimized.

[0030] The graph optimization-based fine alignment module constructs a graph optimization system G based on the map node data to be aligned. match= {V match ,E match}Optimize and adjust the original SLAM data, where: V match =V SLAM ∪V match map , and keep V match map The node attributes in the set are fixed, which is used as one of the optimization benchmarks; match =Relation(V match map ,V match SLAM ∪E SLAM , the output of the previous alignment module and the edges of the SLAM topology map are used as edges of the graph optimization to participate in the optimization process. The optimization process will continue to iterate until the maximum number of iterations I is reached max The optimization error is less than the error threshold Error th Finally, a new set of SLAM topology information maps G′ is output. SLAM = {V′ SLAM ,E′ SLAM}, and then through the inverse process of the logic in the SLAM map data processing module, the topology information map is restored to a new SLAM map containing global coordinates.

[0031] This embodiment is directed to the global map alignment method for the SLAM technology of the above system. First, Figure 2 The prior OSM map and SLAM perception data shown in (a) are used to obtain the corresponding Figure 2 (b) shows the topology map. After node matching, the node information in the SLAM topology map is optimized through graph optimization, and the global position coordinate information in the prior map topology map is assigned to the SLAM node to obtain Figure 2 The effect after alignment in (c).

[0032] Compared with the prior art, this method adds global coordinate information to a SLAM map that only has local information without the need for surveying and mapping.

[0033] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A SLAM map global alignment system based on OSM-like maps, characterized in that: include: A priori OSM-like map processing module, a SLAM map data processing module, a global map coarse alignment module, and a graph-optimization-based fine alignment module, wherein: the priori OSM-like map processing module extracts a priori topological maps with semantic nodes meaningful to driving from the original map data; the SLAM map data processing module obtains SLAM map data through the SLAM algorithm, performs sparse processing and key point extraction, and then generates a SLAM topological information map with key points; the global data coarse alignment module performs coarse alignment processing on the priori topological map and the SLAM topological information map, and filters out topological map node pairs that meet the graph optimization fine alignment conditions; the graph optimization fine alignment module constructs a graph optimization system according to the map node data to be aligned, optimizes and adjusts the initially constructed SLAM topological information map, and obtains an optimized SLAM topological information map, and obtains a SLAM map containing global coordinates after inverse processing by the SLAM map data processing module; The prior topological map G map = {V map , E map }, V map and E map Represent the nodes and edges in the topological map respectively; The SLAM topology information map G SLAM = {V SLAM , E SLAM }, V SLAM and E SLAM Respectively represent the nodes and edges in the SLAM topology map; The original map data refers to: a priori OSM-type road network map, that is, map data in a format similar to OpenStreetMap, including: recording the latitude and longitude coordinates of a certain point in the map and semantic information Nodes, marked as Contains road or contour information Ways composed of Nodes, marked as And store possible related information Relations between each information.

2. The SLAM map global alignment system based on OSM-like maps according to claim 1, characterized in that: The extraction refers to: the prior OSM map processing module analyzes the latitude and longitude coordinates and semantic information of the map midpoint in the original map data and road or contour (such as walls, etc.) information According to TAGS N seg and TAGS W seg Extract vehicle-accessible roads After that, extract the As V map , retaining semantic information and building on the relationship information Relations Topological relationship between each other map , forming a topological semantic map G map = {V map , E map }.

3. The SLAM map global alignment system based on OSM-like maps according to claim 1, characterized in that: The sparse processing refers to: extracting fewer key points from the places without curvature changes in the obtained SLAM map data to achieve the effect of sparse processing and reduce the amount of data.

4. The SLAM map global alignment system based on OSM-like maps according to claim 1, characterized in that: The key point extraction is to select a fixed time length T as a sliding window, take the direction of the first positioning point X0 within T time as the positive direction, fit the positioning points within T time as a curve, and mark the extreme point of the curve as a key point V SLAM , which is the inflection point in the true sense, and sparsely select data according to the curvature to reduce the computational burden of later optimization.

5. The SLAM map global alignment system based on OSM-like maps according to claim 1, characterized in that: The rough alignment process includes: topological information alignment and semantic information alignment; Topology information alignment means: comparing the V map and V SLAM The degree of a node and the structure of a certain level of subtrees expanded with itself as the root node are used as the first level of alignment judgment based on the difference in degree and the similarity of the subtree structure; Semantic information alignment means: map In the case of containing semantic information, the matching V SLAM Determine whether it is possible to have the semantic information; The map node data to be aligned refers to: topological map node pairs (V match map , V match SLAM ), which is the result of rough alignment, where: V match map and V match sLAM The number of elements in the two node sets is the same and they have a one-to-one matching relationship.

6. The SLAM map global alignment system based on OSM-like maps according to claim 1, characterized in that: The optimization adjustment mentioned above refers to: optimizing the system G according to the graph match = {V match , E match }, V match =V SLAM ∪V match map , keep V match map The node attributes in the set are fixed, which is used as one of the optimization benchmarks; match =Relation(V match map , V match SLAM )∪E SLAM , the edges in the SLAM map data obtained by the graph optimization fine alignment module and the SLAM topology information map obtained by the SLAM map data processing module are again used as the input of the graph optimization fine alignment module, and iterative optimization is performed until the maximum number of iterations I is reached max Or the optimization error is less than the error threshold Error th So far, the optimized SLAM topology information map is obtained.

7. The SLAM map global alignment system based on OSM-like maps according to claim 1, characterized in that: The inverse processing refers to: restoring the optimized SLAM topology information map to the original SLAM map format, including the coordinate position changes in the SLAM trajectory and environmental information, and the addition of global coordinate information.

8. A global map alignment method for SLAM technology based on the SLAM map global alignment system based on OSM-like maps as described in any one of claims 1 to 7, characterized in that: Firstly, the prior OSM-like map and SLAM perception data are processed to obtain the corresponding topological map. After node matching, the node information in the SLAM topological map is optimized through graph optimization, and the global position coordinate information in the prior map topological map is assigned to the SLAM nodes.

Images