Hierarchical matching positioning method and equipment for autonomous driving tunnel scenarios

Abstract

The present invention discloses a hierarchical matching positioning method and device for an autonomous driving tunnel scenario. The method is for a scenario in which GNSS fails in a tunnel during autonomous driving, and comprises: structurally organizing, storing and vectorizing the geometric and semantic information of high-precision map positioning elements in the tunnel; performing real-time mapping of surround images collected during vehicle driving through feature coding, depth estimation, perspective conversion, bird's-eye view perspective feature decoding, etc., and performing graph structure encoding on the real-time mapping results; performing similarity measurement calculation on the perceived vectorized mapping results and the stored high-precision map positioning element feature set to obtain a coarse matching positioning result; based on the coarse matching positioning result, intercepting local point cloud data as a source point cloud set; pre-processing the real-time point cloud data collected by the vehicle-mounted sensor as a point cloud set to be registered, performing fine registration on the point cloud set to be registered and the source point cloud set through a pose regressor, and calculating the precise positioning result of the vehicle.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving positioning technology, and in particular to a method and device for GNSS (Global Navigation Satellite System) failure scenario in a tunnel during autonomous driving, which utilizes high-precision map positioning element data and data collected by a camera and a lidar sensor to perform feature encoding and double-layer fusion matching positioning, specifically a layered matching positioning method and device for autonomous driving tunnel scenarios. Background Art

[0002] Positioning, as the foundation of autonomous driving technology, is a necessary prerequisite for vehicles to navigate safely in complex traffic environments. The positioning technologies currently used in the industry are generally inertial combined positioning, wheel speed odometer positioning, and point cloud and visual real-time positioning and mapping. Among them, the inertial combined positioning method relies on the transmission of navigation satellite signals such as Beidou, and the point cloud and visual real-time positioning and mapping method relies on sensors to extract rich feature points in the environment for frame matching. The wheel speed odometer method is simple and cheap, but there are problems with position drift and error accumulation, and it relies on flat, rough ground. However, the actual traffic environment is complex. In scenes such as urban tunnels where there are severe signal obstructions, texture loss, and most ups and downs, the above three technical solutions are difficult to obtain accurate positioning results. Summary of the invention

[0003] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a hierarchical matching positioning method and device for autonomous driving tunnel scenarios. It is aimed at the scenario where GNSS fails in the tunnel during autonomous driving. It uses the data collected by cameras and lidars and combines them with high-precision map information to achieve real-time matching and output of vehicle posture, thereby safeguarding the vehicle's safe navigation decisions and realizing safe autonomous driving in all scenarios and speed ranges.

[0004] According to one aspect of the present invention, a hierarchical matching and positioning method for an autonomous driving tunnel scene is provided, comprising:

[0005] Obtain high-precision map positioning features in the tunnel and perform vector encoding;

[0006] Acquire surround image data and perform real-time mapping and encoding;

[0007] Roughly match the encoded surround image mapping results with the high-precision map positioning features in the tunnel to obtain a rough matching positioning result;

[0008] Based on the coarse matching positioning results, the local point cloud data in the global high-precision map is intercepted as the source point cloud set, and the real-time point cloud set to be registered is obtained according to the data collected by the lidar;

[0009] The point cloud set to be registered is registered with the source point cloud set to obtain a fine matching positioning result.

[0010] As a further technical solution, the high-precision map positioning features in the tunnel are obtained and vectorized and encoded, including:

[0011] Divide the global high-precision map in the tunnel into equal-distance segments;

[0012] Determine the positioning elements of each segment high-precision map;

[0013] The positioning elements in each segment are encoded in graph structure.

[0014] As a further technical solution, the surround image data is acquired and real-time mapping and encoding is performed, including:

[0015] The transmission images from six perspectives, front, back, front left, front right, back left, and back right, are used as input and feature encoded by the backbone network and the neck network to obtain the six-perspective feature map.

[0016] Perform perspective conversion on the obtained six-view feature map and project it into the bird's-eye view space to obtain a bird's-eye view feature map;

[0017] The obtained bird's-eye view feature map is parsed to obtain real-time mapping results.

[0018] As a further technical solution, after obtaining the real-time mapping result, it also includes: encoding the real-time mapping result using a coding method consistent with the positioning features of the high-precision map in the tunnel.

[0019] As a further technical solution, a rough match is performed between the encoded surround image mapping result and the high-precision map positioning features in the tunnel, including: performing a similarity measurement calculation between the encoded surround image mapping result and the high-precision map positioning features in the tunnel, and selecting the area with the highest similarity score as the rough matching positioning result.

[0020] As a further technical solution, a real-time point cloud set to be registered is obtained according to the data collected by the laser radar, and the point cloud set to be registered is registered with the source point cloud set, including: preprocessing the data collected in real time by the vehicle-mounted laser radar as the point cloud set to be registered; calculating the pose relationship between the point cloud set to be registered and the source point cloud set; based on the known pose relationship of the local point cloud high-precision map relative to the global high-precision map coordinate system, the pose matrix of the point cloud set to be registered relative to the global high-precision map is calculated as the fine matching and precise positioning result.

[0021] As a further technical solution, the method further includes:

[0022] Determine whether the vehicle is about to enter a tunnel;

[0023] When it is determined that the vehicle is about to enter a tunnel, determining whether a tunnel entry sign at the tunnel entrance is recognized;

[0024] When a tunnel entry sign is recognized, it is determined that the vehicle has entered a tunnel.

[0025] According to one aspect of the present invention, a hierarchical matching and positioning device for an autonomous driving tunnel scene is provided, comprising:

[0026] High-precision map positioning feature encoding module, used to obtain high-precision map positioning features in the tunnel and perform vectorized encoding;

[0027] Visual perception mapping module, used to obtain surround image data and perform real-time mapping and encoding;

[0028] The coarse matching and positioning module is used to roughly match the encoded surround image mapping results with the high-precision map positioning features in the tunnel to obtain the coarse matching and positioning results;

[0029] The point cloud acquisition module is used to intercept the local point cloud data in the global high-precision map as the source point cloud set based on the coarse matching positioning results, and obtain the real-time point cloud set to be registered based on the data collected by the laser radar;

[0030] The fine matching positioning module is used to align the point cloud set to be registered with the source point cloud set to obtain a fine matching positioning result.

[0031] According to one aspect of the present specification, there is provided an electronic device, comprising: at least one processor, at least one memory and a communication interface; wherein the processor, memory and communication interface communicate with each other; the memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the described method.

[0032] According to one aspect of the present invention, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method described.

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

[0034] (1) The present invention aims at the scenario where GNSS fails in a tunnel during autonomous driving. By first combining high-precision map information with data collected by the camera for coarse matching and positioning, and then combining data collected by the on-board lidar for fine matching and positioning, the present invention achieves real-time matching and output of vehicle posture, thus ensuring safe navigation and decision-making for the vehicle.

[0035] (2) Based on high-precision maps, camera and lidar sensor data, the present invention matches the lane network, traffic signs, traffic signs and other information in the map positioning feature facility layer with the surround view image and point cloud step by step, and performs hierarchical and precise positioning in real time, providing a planning and decision-making basis for autonomous driving vehicles in tunnel scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction is given below to the drawings used in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0037] Figure 1 A flow chart of a method for matching and positioning high-precision map fusion images and point cloud data for an autonomous driving tunnel scene provided by an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of the coding structure of the high-precision map positioning feature map in a tunnel provided by an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of the process of real-time mapping technology of surround view image data provided by an embodiment of the present invention;

[0040] Figure 4 A flowchart of a real-time matching and positioning technology for a point cloud high-precision map and vehicle-mounted point cloud data provided by an embodiment of the present invention;

[0041] Figure 5 A schematic diagram of the structure of a device for matching and positioning high-precision map fusion images and point cloud data for an autonomous driving tunnel scene provided by an embodiment of the present invention;

[0042] Figure 6 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that:

[0044] The terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions, for example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to the steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0045] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices. The flowcharts shown in the accompanying drawings are merely exemplary and do not necessarily include all the contents and operations / steps, nor do they necessarily have to be executed in the order described. For example, some operations / steps may be further decomposed, while some operations / steps may be combined or partially combined, so the actual execution order may change according to the actual situation.

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in 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. In addition, the technical features in the various embodiments or single embodiments provided by the present invention are arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the sequence of steps and / or the structural composition mode, but must be based on the ability of ordinary technicians in this field to achieve. When the combination of technical solutions is contradictory or cannot be achieved, it should be considered that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0047] The embodiment of the present invention provides a hierarchical matching and positioning method for an autonomous driving tunnel scenario. The method integrates high-precision maps with visual and point cloud information to achieve hierarchical matching and positioning. Figure 1 , the method comprising:

[0048] Step 1: vectorize and encode the high-precision map positioning elements in the tunnel;

[0049] Step 2: Real-time mapping of surround view image data is performed using deep learning methods, and the mapping results are encoded in graph structure. Step 3: Similarity measurement is performed between the results of real-time mapping and graph structure encoding of the surround view camera and the vectorized encoded tunnel high-precision map positioning feature set, and a rough match is retrieved to obtain the approximate location interval of the vehicle.

[0050] Step 4, intercepting the local point cloud data within the coarse matching positioning interval in the global high-precision map as the source point cloud set;

[0051] Step 5: Preprocess the vehicle-mounted laser radar point cloud data by denoising, downsampling, dynamic target filtering, etc. as a point cloud set to be registered, and then align the point cloud set to be registered with the source point cloud set to obtain a fine matching positioning result.

[0052] It should be noted that step 2 uses deep learning methods to learn and optimize network parameters to output ideal results. In the algorithms involved in real-time mapping, such as feature encoding, depth estimation, perspective conversion, and bird's-eye view perspective feature decoding, each part involves parameter learning.

[0053] Based on the content of the above method embodiment, as a preferred embodiment, a hierarchical matching positioning method for autonomous driving tunnel scenes is provided in the embodiments of the present invention. When the high-precision map positioning elements in the tunnel are vectorized and encoded with a graph structure, the high-precision map positioning elements include: high-precision vector lane network, traffic facilities and other element geometry and attribute information. In particular, in addition to the common lane arrows, the traffic facilities in the tunnel scene also include tunnel signs, tunnel entry warning signs, tunnel length signs, tunnel speed limit signs, tunnel driving light signs, speed limit release signs, safety doors, milestones, etc. These elements constitute a structured combined positioning feature with uniqueness and reliability. At the same time, in order to perform similarity measurement calculations with the real-time perception mapping results of sensor data, the embodiments of the present invention use a graph structure to vectorize the combined positioning features.

[0054] Preferably, see Figure 2 The method includes: selecting the geometry and attribute information of elements such as lane lines, lane arrows, tunnel signs, tunnel entry warning signs, tunnel length signs, tunnel speed limit signs, tunnel headlight signs, speed limit release signs, safety doors, and milestones in the tunnel as positioning elements, and segmenting the global high-precision map in the tunnel, and performing graph structure encoding on the positioning elements in each segment to form a high-precision tunnel map positioning feature set.

[0055] The specific encoding method is as follows: for linear elements such as lane lines, point interpolation is performed uniformly according to the distance, and the interpolation point set is used as the node. The adjacency matrix is ​​constructed according to the topological relationship between the lines, and each line element is regarded as an independent subgraph; for the traffic facilities on the roadside or above the tunnel, its minimum circumscribed rectangle is calculated, and the nodes of the rectangle are selected by the equidistant uniform interpolation method, and the adjacency matrix is ​​constructed according to the node connection relationship. Each traffic facility is regarded as an independent subgraph; at the same time, the adjacency matrix is ​​constructed between subgraphs based on the topological relationship; finally, each segmented area map is vectorized and encoded through a graph neural network.

[0056] Based on the content of the above method embodiment, as a preferred embodiment, a hierarchical matching and positioning method for an autonomous driving tunnel scene is provided in the embodiment of the present invention, wherein the surround image data is mapped in real time by a learning method and a graph structure is constructed, see Figure 3, the method includes: first, using the perspective images of six perspectives, namely, front, back, front left, front right, back left, and back right, as input, and respectively passing through the backbone network (such as resnet, Swin Transformer, etc.) and the neck network (FPN) for feature encoding to obtain the six-perspective feature map, performing depth estimation on the six-perspective feature map, and using the camera internal and external parameter matrix to convert the six-perspective feature map, and then projecting it into the bird's-eye view space to form a high-dimensional bird's-eye view feature map. Then, the bird's-eye view features are parsed by the Transformer decoder, and the high-precision map elements detected by the mapping detection head are output as real-time mapping results. Finally, the real-time mapping results are constructed using the graph structure encoding method consistent with the above-mentioned high-precision map positioning features. The mapping detection head refers to the head layer (MLP layer) of the deep learning mapping network, which is used to output high-precision map detection results.

[0057] In the method provided in this embodiment, the vehicle-mounted sensor includes a six-view camera. The three-dimensional scene map is constructed based on the image data of the six-view camera through feature encoding, depth estimation, perspective conversion, bird's-eye view perspective feature decoding and other algorithms, and the real-time mapping result is encoded with the graph structure using the method in step 1.

[0058] Based on the content of the above method embodiment, as a preferred embodiment, a hierarchical matching and positioning method for autonomous driving tunnel scenarios is provided in the embodiments of the present invention. The surround image mapping results are roughly matched with the vectorized tunnel high-precision map positioning feature set to obtain the approximate position range of the vehicle. The method includes: performing similarity measurement calculation on the high-precision map elements that are perceived in real time and encoded with the graph structure and the segmented high-precision map positioning elements that are also encoded with the graph structure, and selecting the area with the highest similarity score as the coarse matching positioning result.

[0059] Based on the content of the above method embodiment, as a preferred embodiment, a hierarchical matching and positioning method for an autonomous driving tunnel scene is provided in the embodiment of the present invention, wherein the local point cloud data within the coarse matching and positioning interval in the global high-precision map is intercepted as a source point cloud set, including: using the above coarse matching and positioning interval as a constraint space, intercepting the point cloud of the local interval as a source point cloud set. It should be noted that the storage form of the high-precision map includes both the above-mentioned structured points, lines, surfaces and other elements, and point cloud sets.

[0060] As a preferred embodiment, the local point cloud data in the rough matching positioning interval of the global high-precision map provided in the embodiment of the present invention is intercepted as the source point cloud set, and the vehicle-mounted laser radar point cloud data is pre-processed by noise reduction, downsampling, dynamic target filtering, etc. as the point cloud set to be registered, and then the point cloud set to be registered is registered with the source point cloud set to obtain a fine positioning result, see Figure 4The method includes: based on the coarse matching positioning result obtained by the above method, intercepting the high-precision map point cloud data in the coarse matching positioning area as the source point cloud set, and performing pre-processing processes such as denoising, downsampling, and dynamic target filtering on the data collected in real time by the vehicle-mounted laser radar as the point cloud set to be registered, matching the point cloud set to be registered with the source point cloud set, calculating the relative pose relationship, and then calculating the pose matrix between the point cloud set to be registered and the global high-precision map based on the known pose of the local point cloud high-precision map relative to the global high-precision map coordinate system as the fine matching positioning result.

[0061] Furthermore, after obtaining the point cloud set to be registered, the key points, feature descriptors and other information of the two point cloud sets are combined, and the two point cloud data sets are registered by calculating the rigid body transformation matrix between the point cloud set to be registered and the source point cloud, and finally an accurate fine matching positioning result is obtained.

[0062] In the method provided in this embodiment, the vehicle-mounted sensor includes a laser radar. After pre-processing processes such as denoising, downsampling, and dynamic target filtering are performed on the laser radar point cloud data, a point cloud set to be registered is obtained.

[0063] Based on the content of the above method embodiment, as a preferred embodiment, a hierarchical matching and positioning method for an autonomous driving tunnel scene provided in an embodiment of the present invention is described in detail in detail in FIG. Figure 1 , and also includes: judging whether to enter a tunnel by using the positioning result of the GNSS inertial navigation system, and then judging whether to enter the tunnel by sensing and identifying the tunnel entry sign at the tunnel entrance.

[0064] The implementation basis of each embodiment of the present invention is to implement programmed processing through a device with a processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this reality, on the basis of the above embodiments, an embodiment of the present invention provides a hierarchical matching and positioning device for an autonomous driving tunnel scene, which is used to execute the hierarchical matching and positioning method for an autonomous driving tunnel scene in the above method embodiment.

[0065] See also Figure 5The device includes: a high-precision map positioning feature encoding module, which is used to obtain the high-precision map positioning features in the tunnel and perform vectorized encoding; a visual perception mapping module, which is used to obtain surround image data and perform real-time mapping and encoding; a coarse matching positioning module, which is used to coarsely match the encoded surround image mapping results with the high-precision map positioning features in the tunnel to obtain a coarse matching positioning result; a point cloud acquisition module, which is used to intercept local point cloud data in the global high-precision map as a source point cloud set based on the coarse matching positioning result, and obtain a real-time point cloud set to be registered based on the data collected by the laser radar; a fine matching positioning module, which is used to align the point cloud set to be registered with the source point cloud set to obtain a fine matching positioning result.

[0066] The hierarchical matching positioning device for the autonomous driving tunnel scenario provided by the embodiment of the present invention is designed for the scenario where the tunnel GNSS positioning system fails during the autonomous driving process. Figure 5 Several modules in it use the positioning features of high-precision maps and the data collected by cameras and lidar sensors to achieve lane-level precise positioning through processes such as feature encoding, coarse matching, and fine matching, thereby ensuring safe driving in all scenarios and all speed ranges for autonomous driving.

[0067] It should be noted that the device embodiments provided by the present invention are used to implement the methods in the above method embodiments as well as the methods in other method embodiments provided by the present invention. The only difference is that the corresponding functional modules are set, and the principle is basically the same as the principle of the above device embodiments provided by the present invention. As long as the technical personnel in the field refer to the specific technical solutions in other method embodiments on the basis of the above device embodiments, obtain the corresponding technical means and the technical solutions composed of these technical means by combining the technical features, and on the premise of ensuring the practicality of the technical solutions, they can improve the devices in the above device embodiments to obtain the corresponding device class embodiments, which are used to implement the methods in other method class embodiments. For example:

[0068] Based on the content of the above device embodiment, as a preferred embodiment, the hierarchical matching and positioning device for the autonomous driving tunnel scene provided in the embodiment of the present invention further includes:

[0069] The first high-precision map submodule is used to divide the global high-precision map in the tunnel into equal-distance segments;

[0070] The second high-precision map submodule is used to determine the positioning elements of each segment high-precision map;

[0071] The third high-precision map sub-module is used to perform graph structure encoding on the positioning elements in each segment.

[0072] Based on the content of the above device embodiment, as a preferred embodiment, the hierarchical matching and positioning device for the autonomous driving tunnel scene provided in the embodiment of the present invention further includes:

[0073] The first real-time mapping submodule is used to take the transmission images of six perspectives, namely, front, back, front left, front right, back left, and back right, as input, and perform feature encoding through the backbone network and the neck network respectively to obtain the six-perspective feature map;

[0074] The second real-time mapping submodule is used to perform perspective conversion on the obtained six-view feature map and project it into the bird's-eye view space to obtain a bird's-eye view feature map;

[0075] The third real-time mapping submodule is used to analyze the obtained bird's-eye view feature map to obtain a real-time mapping result.

[0076] Based on the content of the above device embodiment, as a preferred embodiment, the hierarchical matching and positioning device for the autonomous driving tunnel scene provided in the embodiment of the present invention further includes:

[0077] The fourth real-time mapping submodule is used to encode the real-time mapping results using a coding method consistent with the positioning features of the high-precision map in the tunnel.

[0078] Based on the content of the above device embodiment, as a preferred embodiment, the hierarchical matching and positioning device for the autonomous driving tunnel scene provided in the embodiment of the present invention further includes:

[0079] The first coarse matching submodule is used to calculate the similarity between the encoded surround image mapping result and the positioning features of the high-precision map in the tunnel;

[0080] The second coarse matching submodule is used to select the area with the highest similarity score as the coarse matching positioning result.

[0081] Based on the content of the above device embodiment, as a preferred embodiment, the hierarchical matching and positioning device for the autonomous driving tunnel scene provided in the embodiment of the present invention further includes:

[0082] The first fine matching submodule is used to pre-process the data collected in real time by the vehicle-mounted laser radar as a point cloud set to be registered;

[0083] The second fine matching submodule is used to calculate the pose relationship between the point cloud set to be registered and the source point cloud set;

[0084] The third fine matching submodule is used to calculate the pose matrix of the point cloud set to be registered relative to the global high-precision map based on the known pose relationship of the local point cloud high-precision map relative to the global high-precision map coordinate system, as the fine matching precise positioning result.

[0085] Based on the content of the above device embodiment, as a preferred embodiment, the hierarchical matching and positioning device for the autonomous driving tunnel scene provided in the embodiment of the present invention further includes:

[0086] The first judgment submodule is used to judge whether the vehicle is about to enter a tunnel;

[0087] The second judgment submodule is used to judge whether the vehicle enters the tunnel.

[0088] The method of the embodiment of the present invention is implemented by relying on electronic devices, so it is necessary to introduce the relevant electronic devices. Based on this purpose, the embodiment of the present invention provides an electronic device, such as Figure 6 As shown, the electronic device includes: at least one processor, a communication interface, at least one memory, and a communication bus, wherein at least one processor, the communication interface, and at least one memory communicate with each other through the communication bus. At least one processor calls the logic instructions in at least one memory to execute all or part of the steps of the method provided by the aforementioned various method embodiments.

[0089] In addition, when the logic instructions in the above-mentioned at least one memory are implemented in the form of a software functional unit and sold or used as an independent product, they are stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or partly in other words, the part that contributes to the prior art or part of the technical solution is embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for enabling a computer device (a personal computer, a server, or a network device) to perform all or part of the steps of the method described in each method embodiment of the present invention. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, and various media for storing program codes.

[0090] The system embodiments described above are merely illustrative, wherein the units described as separate components are or are not physically separated, and the components shown as units are or are not physical units, located in one place, or distributed on multiple network units. Some or all of the modules are selected according to actual conditions to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement without creative effort.

[0091] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0093] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0095] In summary, the present invention provides a high-precision map fusion perception layered matching positioning method for autonomous driving in a tunnel GNSS failure scenario. The method includes: structurally organizing, storing and vectorizing the geometric and semantic information of high-precision map positioning feature elements such as lane lines in tunnel GNSS failure scenarios and special traffic signs in tunnels; real-time mapping of the surround images collected by sensors during vehicle driving through algorithms such as feature coding, depth estimation, perspective conversion, bird's-eye view perspective feature decoding, and image structure encoding of the real-time mapping results; similarity measurement calculation of the perceived mapping results and the stored high-precision map positioning feature element feature set, and rough matching to obtain a rough positioning interval; based on the rough matching positioning results, intercepting the local point cloud data of the corresponding area in the global high-precision map as the source point cloud set; pre-processing the real-time point cloud data collected by the vehicle-mounted sensor such as denoising, downsampling, dynamic target filtering, etc. as the point cloud set to be registered, and finely aligning the point cloud set to be registered with the source point cloud set through a pose regressor to calculate the precise positioning result of the vehicle. The present invention is aimed at the scenario where the GNSS positioning system in the tunnel fails during autonomous driving. It uses the positioning features of high-precision maps and the data collected by cameras and lidar sensors to achieve lane-level precise positioning through processes such as feature encoding, coarse matching, and fine matching, thereby ensuring safe driving in all scenarios and all speed ranges of autonomous driving.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A hierarchical matching and positioning method for autonomous driving tunnel scenarios, characterized in that: include: Obtain high-precision map positioning features in the tunnel and perform vector encoding; Acquire surround image data and perform real-time mapping and encoding; The similarity measurement calculation is performed on the mapping results of the encoded surround view image and the positioning features of the high-precision map in the tunnel, and the area with the highest similarity score is selected as the coarse matching positioning result; Based on the coarse matching positioning results, the local point cloud data in the global high-precision map is intercepted as the source point cloud set, and the data collected in real time by the vehicle-mounted laser radar is preprocessed as the point cloud set to be registered; Calculate the pose relationship between the point cloud set to be registered and the source point cloud set; based on the known pose relationship of the local point cloud high-precision map relative to the global high-precision map coordinate system, calculate the pose matrix of the point cloud set to be registered relative to the global high-precision map as the fine matching and precise positioning result.

2. The hierarchical matching and positioning method for autonomous driving tunnel scenes according to claim 1 is characterized in that: Obtain high-precision map positioning features in the tunnel and perform vector coding, including: Divide the global high-precision map in the tunnel into equal-distance segments; Determine the positioning elements of each segment high-precision map; The positioning elements in each segment are encoded in graph structure.

3. The hierarchical matching and positioning method for autonomous driving tunnel scenes according to claim 1 is characterized in that: Acquire surround image data and perform real-time mapping and encoding, including: The transmission images from six perspectives, front, back, front left, front right, back left, and back right, are used as input and feature encoded by the backbone network and the neck network to obtain the six-perspective feature map. Perform perspective conversion on the obtained six-view feature map and project it into the bird's-eye view space to obtain a bird's-eye view feature map; The obtained bird's-eye view feature map is parsed to obtain real-time mapping results.

4. The hierarchical matching and positioning method for autonomous driving tunnel scenes according to claim 3 is characterized in that: After obtaining the real-time mapping result, the method further includes: encoding the real-time mapping result using a coding method consistent with the positioning features of the high-precision map in the tunnel.

5. The hierarchical matching and positioning method for autonomous driving tunnel scenes according to claim 1 is characterized in that: The method further comprises: Determine whether the vehicle is about to enter a tunnel; When it is determined that the vehicle is about to enter a tunnel, determining whether a tunnel entry sign at the tunnel entrance is recognized; When a tunnel entry sign is recognized, it is determined that the vehicle has entered a tunnel.

6. A hierarchical matching and positioning device for autonomous driving tunnel scenarios, characterized in that: include: High-precision map positioning feature encoding module, used to obtain high-precision map positioning features in the tunnel and perform vectorized encoding; Visual perception mapping module, used to obtain surround image data and perform real-time mapping and encoding; The coarse matching and positioning module is used to calculate the similarity between the encoded surround image mapping results and the high-precision map positioning features in the tunnel, and select the area with the highest similarity score as the coarse matching and positioning result; The fine matching and positioning module is used to pre-process the data collected in real time by the vehicle-mounted laser radar as a point cloud set to be registered; Calculate the pose relationship between the point cloud set to be registered and the source point cloud set; based on the known pose relationship of the local point cloud high-precision map relative to the global high-precision map coordinate system, calculate the pose matrix of the point cloud set to be registered relative to the global high-precision map as the fine matching and precise positioning result.

7. An electronic device, characterized in that: include: At least one processor, at least one memory and a communication interface; wherein the processor, memory and communication interface communicate with each other; The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the method according to any one of claims 1 to 5.

8. A non-transitory computer readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which cause the computer to execute the method of any one of claims 1 to 5.

Images