Point cloud-based high-precision map updating methods, devices, vehicles, and media

Abstract

This application relates to the field of high-definition mapping, and particularly to a method, apparatus, vehicle, and medium for updating high-definition maps based on point clouds. The method includes: acquiring point cloud data, pose data, and image data of the vehicle's current location; obtaining a local point cloud map based on the pose data and a preset point cloud map, and obtaining an optimal matching pose result that meets preset accuracy conditions based on the local point cloud map and the point cloud data; and updating the current local high-definition map based on the optimal matching pose result and the image data. Therefore, by updating a local high-definition map using point cloud data, pose data, and image data, this method solves the problems of error variations caused by multiple vehicles collecting data on the same road segment or multiple collections by a single vehicle, leading to decreased global consistency of the high-definition map and abrupt changes in high-definition map elements. It achieves automatic updating of elements in the high-definition map without misalignment, map blurring, or overlay.

Description

Technical Field

[0001] This application relates to the field of high-precision mapping, and in particular to a method, apparatus, vehicle, and medium for updating high-precision maps based on point clouds. Background Technology

[0002] High-precision maps, also known as high-resolution maps, possess accurate vehicle location information and rich road element data. This helps vehicles anticipate complex road conditions such as slope, curvature, and heading, allowing them to better avoid potential driving risks. Compared to ordinary electronic maps, high-precision maps have a larger data scale, thus requiring higher processing performance and efficiency during updates to ensure timeliness, accuracy, and reliability.

[0003] In related technologies, the use and updating of high-precision maps are usually achieved by collecting road segment data through vehicle sensors.

[0004] However, when the same road segment is collected multiple times by the same vehicle or by different vehicles in turn, the detected map features may be misaligned due to sensor errors and noise. If the collected data is directly updated to the map, it will lead to a decrease in the global consistency of the high-precision map and sudden changes in the high-precision map features, which urgently needs to be addressed. Summary of the Invention

[0005] This application provides a high-precision map update method, device, vehicle, and medium based on point cloud to solve the problems of error changes caused by multiple vehicles collecting data on the same road segment or a single vehicle collecting data multiple times in the current update method, which leads to a decrease in the global consistency of the high-precision map and sudden changes in high-precision map elements. It realizes the automatic update of elements in the high-precision map without misalignment, map blurring, or overlay.

[0006] To achieve the above objectives, the first aspect of this application proposes a high-precision map update method based on point clouds, comprising the following steps:

[0007] Acquire point cloud data, pose data, and image data of the vehicle's current location;

[0008] A local point cloud map is obtained based on the pose data and a preset point cloud map, and an optimal matching pose result satisfying a preset accuracy condition is obtained based on the local point cloud map and the point cloud data; and

[0009] Based on the optimal matching pose result, the current local high-precision map is updated according to the image data.

[0010] According to one embodiment of this application, obtaining the local point cloud map based on the pose data and the preset point cloud map includes:

[0011] Obtain the keyframe sequence in the preset point cloud map, wherein the keyframe sequence includes the number and pose of each keyframe;

[0012] Construct a Kd tree based on the keyframe sequence, and determine the keyframe number that has the smallest pose distance from the pose data from the Kd tree;

[0013] The local point cloud map is obtained by stitching together the minimum keyframe number in the first preset frame data in the first direction and the minimum keyframe number in the second preset frame data in the second direction.

[0014] According to one embodiment of this application, obtaining the optimal matching pose result that satisfies the preset accuracy condition based on the local point cloud map and the point cloud data includes:

[0015] Based on the local point cloud map, the point cloud data is optimized using a preset ICP (Iterative Closest Point) algorithm to obtain the optimal matching pose result and the matching score corresponding to the optimal matching pose result;

[0016] If the matching score is greater than or equal to the preset score, then the optimal matching pose result is determined to meet the preset accuracy condition.

[0017] According to one embodiment of this application, after obtaining the optimal matching pose result and the matching score corresponding to the optimal matching pose result, the method further includes:

[0018] If the matching score is less than the preset score, the optimal matching pose result is determined to not meet the preset accuracy condition, and the update of the current local high-precision map is stopped.

[0019] According to one embodiment of this application, updating the current local high-precision map based on the image data according to the optimal matching pose result includes:

[0020] The current frame image is processed using a preset deep learning algorithm to obtain multiple map elements;

[0021] Take any one of the multiple map elements as the current map element, and determine whether there is a target map element in the current local high-precision map that corresponds to the current map element;

[0022] If there is no target map element corresponding to the current map element in the current local high-definition map, then the current map element is added to the current local high-definition map; otherwise, if the matching score of the current map element is greater than the matching score of the target map element, then the current map element replaces the target map element; if the matching score of the current map element is less than or equal to the matching score of the target map element, then the current local high-definition map is not updated; this process continues until all the map elements have been traversed.

[0023] The high-precision map updating method based on point cloud proposed in this application obtains a local point cloud map by using the pose data of the vehicle's current location and a preset point cloud map. Then, based on the local point cloud map and the point cloud data of the vehicle's current location, an optimal matching pose result satisfying preset accuracy conditions is obtained. Based on the optimal matching pose result, the current local high-precision map is updated using image data. Therefore, by updating the local high-definition map using point cloud data, pose data, and image data, this method solves the problems of error variations caused by multiple vehicles collecting data on the same road segment or multiple collections by a single vehicle in current updating methods, leading to decreased global consistency of the high-precision map and abrupt changes in high-precision map elements. This achieves automatic updating of elements in the high-precision map without misalignment, map blurring, or overlay.

[0024] To achieve the above objectives, a second aspect of this application provides a high-precision map updating device based on point clouds, comprising:

[0025] The acquisition module is used to acquire point cloud data, pose data, and image data of the vehicle's current location;

[0026] The processing module is configured to obtain a local point cloud map based on the pose data and a preset point cloud map, and to obtain an optimal matching pose result that meets preset accuracy conditions based on the local point cloud map and the point cloud data; and

[0027] The update module is used to update the current local high-precision map based on the optimal matching pose result and the image data.

[0028] According to one embodiment of this application, the processing module includes:

[0029] The acquisition unit is used to acquire the key frame sequence in the preset point cloud map, wherein the key frame sequence includes the number and pose of each key frame;

[0030] A construction unit is configured to construct a Kd tree based on the key frame sequence and determine the key frame number that has the smallest pose distance from the pose data from the Kd tree.

[0031] The stitching unit is used to stitch together the minimum keyframe number in the first direction and the minimum keyframe number in the second direction to obtain the local point cloud map.

[0032] According to one embodiment of this application, the processing module includes:

[0033] The optimization unit is used to optimize the point cloud data based on the local point cloud map using a preset ICP algorithm to obtain the optimal matching pose result and the matching score corresponding to the optimal matching pose result.

[0034] The determination unit is used to determine that the optimal matching pose result satisfies the preset accuracy condition when the matching score is greater than or equal to a preset score.

[0035] According to one embodiment of this application, after obtaining the optimal matching pose result and the matching score corresponding to the optimal matching pose result, the optimization unit is further configured to:

[0036] If the matching score is less than the preset score, the optimal matching pose result is determined to not meet the preset accuracy condition, and the update of the current local high-precision map is stopped.

[0037] According to one embodiment of this application, the updating module is specifically used for:

[0038] The current frame image is processed using a preset deep learning algorithm to obtain multiple map elements;

[0039] Take any one of the multiple map elements as the current map element, and determine whether there is a target map element in the current local high-precision map that corresponds to the current map element;

[0040] If there is no target map element corresponding to the current map element in the current local high-definition map, then the current map element is added to the current local high-definition map; otherwise, if the matching score of the current map element is greater than the matching score of the target map element, then the current map element replaces the target map element; if the matching score of the current map element is less than or equal to the matching score of the target map element, then the current local high-definition map is not updated; this process continues until all the map elements have been traversed.

[0041] The high-precision map updating device based on point cloud proposed in this application obtains a local point cloud map by using the pose data of the vehicle's current location and a preset point cloud map. Then, it obtains an optimal matching pose result that meets preset accuracy conditions based on the local point cloud map and the point cloud data of the vehicle's current location. Based on the optimal matching pose result, the current local high-precision map is updated using image data. Therefore, by updating the local high-definition map using point cloud data, pose data, and image data, this solves the problems of error variations caused by multiple vehicles collecting data on the same road segment or multiple collections by a single vehicle in current updating methods, leading to decreased global consistency of the high-precision map and sudden changes in high-precision map elements. This achieves automatic updating of elements in the high-precision map without misalignment, map blurring, or overlay.

[0042] To achieve the above objectives, a third aspect of this application provides a vehicle comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the point cloud-based high-precision map update method as described in the above embodiments.

[0043] To achieve the above objectives, the fourth aspect of this application provides a computer storage medium storing a computer program that is executed by a processor to implement the point cloud-based high-precision map update method as described in the above embodiments.

[0044] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0045] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0046] Figure 1 This is a flowchart of a high-precision map update method based on point clouds provided according to an embodiment of this application;

[0047] Figure 2 This is a schematic diagram illustrating the correspondence between point cloud maps and high-precision maps according to an embodiment of this application;

[0048] Figure 3 This is a data flow diagram for local high-precision map updating according to an embodiment of this application;

[0049] Figure 4 This is a flowchart of a point cloud-based high-precision map update method according to another embodiment of this application;

[0050] Figure 5This is a block diagram of a point cloud-based high-precision map updating device according to an embodiment of this application;

[0051] Figure 6 This is a structural schematic diagram of a vehicle according to an embodiment of this application. Detailed Implementation

[0052] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0053] The following describes, with reference to the accompanying drawings, a method, apparatus, vehicle, and medium for updating high-precision maps based on point clouds according to embodiments of this application. First, the method for updating high-precision maps based on point clouds according to embodiments of this application will be described with reference to the accompanying drawings.

[0054] Figure 1 This is a flowchart of a point cloud-based high-precision map update method according to an embodiment of this application.

[0055] like Figure 1 As shown, this point cloud-based high-precision map update method includes the following steps:

[0056] In step S101, point cloud data, pose data and image data of the vehicle's current location are acquired.

[0057] Point cloud data can be collected by lidar. The lidar system scans the ground to obtain the three-dimensional coordinates of ground reflection points. Each ground reflection point is distributed in three-dimensional space as a point according to the three-dimensional coordinates, which is called a scan point. Point cloud data refers to the collection of scan points in a three-dimensional coordinate system. Pose data can be obtained by position sensors. Image data (such as lane lines, traffic signs, etc.) can be obtained by image sensors.

[0058] In step S102, a local point cloud map is obtained based on the pose data and the preset point cloud map, and the optimal matching pose result that meets the preset accuracy conditions is obtained based on the local point cloud map and the point cloud data.

[0059] Point cloud map is a type of map information acquired by sensors such as radar and stored in the form of point cloud data. It is used to describe a three-dimensional model of objects and environment on the ground. The preset point cloud map is the global point cloud map after the last update. In this embodiment, a local point cloud map can be obtained based on the pose data and the preset point cloud map. Then, an optimized (i.e., satisfying the preset accuracy conditions) optimal matching pose result can be obtained based on the local point cloud map and the point cloud data.

[0060] Understandably, vehicle-mounted LiDAR has high measurement accuracy, a large measurement range, and is not affected by lighting conditions, resulting in high accuracy in point cloud matching. Therefore, updating subsequent local high-precision maps based on a complete global point cloud map is more accurate and reliable.

[0061] Furthermore, in some embodiments, obtaining a local point cloud map based on pose data and a preset point cloud map includes: acquiring a keyframe sequence in the preset point cloud map, wherein the keyframe sequence includes the number and pose of each keyframe; constructing a Kd-tree based on the keyframe sequence, and determining the keyframe number with the smallest pose distance from the pose data from the Kd-tree; and stitching together a first preset number of frame data in a first direction and a second preset number of frame data in a second direction based on the smallest keyframe number to obtain the local point cloud map.

[0062] The first preset frame number and the second preset frame number can be preset by those skilled in the art, obtained through a limited number of experiments, or obtained through a limited number of computer simulations; no specific limitation is made here.

[0063] Specifically, in this embodiment, the keyframe sequence in the preset point cloud map can be obtained through the SLAM (Simultaneous Localization and Mapping) algorithm. The keyframe sequence includes the number of each keyframe and its corresponding pose, which can be represented as {(Cloud1,Pose1)(Cloud2,Pose2),...,(Cloudm,Posem)}, where Cloudm represents the number of the m-th keyframe and Posem represents the pose of the m-th keyframe. After obtaining the keyframe sequence, a Kd-tree (short for K-dimensional tree, a data structure for segmenting k-dimensional data space) can be constructed based on the keyframe sequence. Based on the pose data obtained in step S101, the keyframe number with the smallest pose distance in the pose data is determined from the Kd-tree. The first preset number of frames (can be labeled, such as m frames) of the smallest keyframe number in the first direction (e.g., forward) and the second preset number of frames (can be labeled, such as m frames) of the smallest keyframe number in the second direction (e.g., backward) are then concatenated to obtain the local point cloud map.

[0064] Furthermore, in some embodiments, obtaining the optimal matching pose result that meets the preset accuracy condition based on the local point cloud map and point cloud data includes: optimizing the point cloud data using a preset ICP algorithm based on the local point cloud map to obtain the optimal matching pose result and the matching score corresponding to the optimal matching pose result; if the matching score is greater than or equal to a preset score, then it is determined that the optimal matching pose result meets the preset accuracy condition.

[0065] Specifically, in this embodiment, the point cloud data obtained in step S101 can be optimized based on a local point cloud map using a preset ICP algorithm. That is, the point cloud data and the local point cloud map are corrected to obtain the optimized optimal matching pose result and the matching score corresponding to the optimal matching pose result. When the matching score is greater than or equal to a preset score, it can be determined that the optimal matching pose result meets the preset accuracy condition, and the current local high-precision map can continue to be updated.

[0066] In step S103, the current local high-precision map is updated based on the optimal matching pose result and the image data.

[0067] Furthermore, in some embodiments, based on the optimal matching pose result, the current local high-precision map is updated according to the image data, including: processing the current frame image using a preset deep learning algorithm to obtain multiple map elements; taking any one of the multiple map elements as the current map element, and determining whether there is a target map element corresponding to the current map element in the current local high-precision map; if there is no target map element corresponding to the current map element in the current local high-precision map, then adding the current map element to the current local high-precision map; otherwise, when the matching score of the current map element is greater than the matching score of the target map element, replacing the target map element with the current map element; when the matching score of the current map element is less than or equal to the matching score of the target map element, not updating the current local high-precision map; until multiple map elements are traversed.

[0068] Among them, the current local high-precision map can extract road and environmental elements (such as lane lines, traffic signs, etc.) from single point cloud data or multiple fused point cloud data through a preset deep learning algorithm, save the type, location and shape of these elements, and then manually or automatically annotate them based on the local point cloud map.

[0069] Specifically, in this embodiment, a preset deep learning algorithm is used to process the current frame image data to obtain multiple map elements. Any map element is selected from these multiple map elements as the current map element Mi. Based on the position of the current map element Mi, it is determined whether a target map element Oj corresponding to the current map element Mi exists in the current local high-definition map. If no target map element Oj corresponding to the current map element Mi exists in the current local high-definition map, the current map element Mi can be added to the current local high-definition map. If a corresponding target map element Oj exists in the current local high-definition map, the matching scores of the current map element Mi and the corresponding target map element Oj are compared. If the matching score of the current map element Mi is greater than the matching score of the corresponding target map element Oj, then the current map element Mi replaces the corresponding target map element Oj. If the matching score of the current map element Mi is less than or equal to the matching score of the corresponding target map element Oj, then the current local high-definition map is not updated. This logic continues until each map element is traversed.

[0070] In addition, in some embodiments, after obtaining the optimal matching pose result and the matching score corresponding to the optimal matching pose result, the method further includes: if the matching score is less than a preset score, then it is determined that the optimal matching pose result does not meet the preset accuracy condition, and the update of the current local high-precision map is stopped.

[0071] It is understandable that after obtaining the optimal matching pose result and the matching score corresponding to the optimal matching pose result, if the matching score is not within a reasonable range, that is, the matching score is less than the preset score, then it is determined that the optimal matching pose result does not meet the preset accuracy condition, and the point cloud data obtained in step S101 can be discarded directly, and the update of the current local high-precision map can be stopped.

[0072] It should be noted that, as Figure 2 As shown, the local point cloud map corresponding to the current local high-precision map must be complete, that is, each map element in the current local high-precision map can find a corresponding keyframe in the local point cloud map, and the same keyframe may correspond to different map elements.

[0073] Furthermore, such as Figure 3 As shown, Figure 3 The data flow diagram for updating the current high-precision map is as follows: Step S101 obtains point cloud data and pose data, and matches them with a preset point cloud map to obtain the optimal matching pose result relative to the preset point cloud map, which is equivalent to transforming from a global coordinate system to a local coordinate system; Step S101 obtains point cloud data and image data, and multiple map elements (such as lane lines, traffic signs, etc.) can be extracted from them through a preset deep learning algorithm; Based on the optimal matching pose result and the map elements in the image data, the current local high-precision map is updated.

[0074] To facilitate those skilled in the art to further understand the point cloud-based high-precision map update method proposed in the embodiments of this application, the following is combined with... Figure 4 To provide further explanation.

[0075] like Figure 4 As shown, another embodiment of the high-precision map update method based on point clouds in this application includes the following steps:

[0076] Step S401: Acquire point cloud data, pose data, and graphic data.

[0077] Step S402: Search for the keyframe number with the smallest distance from the pose data in the preset point cloud map, and stitch together the keyframe data of m frames before and after the smallest keyframe number to obtain a local point cloud map.

[0078] Step S403: The point cloud data and the local point cloud map are used to obtain the optimal matching pose result and the matching score corresponding to the optimal matching pose result through the ICP algorithm. If the matching score is less than the preset score, the update is terminated.

[0079] Step S404: Traverse the multiple map elements Mi obtained from the current frame image.

[0080] Step S405: Determine whether a target map element Oj corresponding to Mi exists in the current local high-precision map based on the position of Mi. If it exists, proceed to step S406; otherwise, proceed to step S409.

[0081] Step S406: Determine whether the matching score of Mi is greater than the matching score of Oj. If it is greater, proceed to step S407; otherwise, proceed to step S408.

[0082] Step S407: Delete Oj and replace Oj with Mi in the current local high-precision map.

[0083] Step S408: Do not update the current local high-precision map.

[0084] Step S409: Add Mi to the current local high-precision map, and record the matching score and keyframe number of Mi.

[0085] The high-precision map updating method based on point cloud proposed in this application obtains a local point cloud map by using the pose data of the vehicle's current location and a preset point cloud map. Then, based on the local point cloud map and the point cloud data of the vehicle's current location, an optimal matching pose result satisfying preset accuracy conditions is obtained. Based on the optimal matching pose result, the current local high-precision map is updated using image data. Therefore, by updating the local high-definition map using point cloud data, pose data, and image data, this method solves the problems of error variations caused by multiple vehicles collecting data on the same road segment or multiple collections by a single vehicle in current updating methods, leading to decreased global consistency of the high-precision map and abrupt changes in high-precision map elements. This achieves automatic updating of elements in the high-precision map without misalignment, map blurring, or overlay.

[0086] Next, referring to the accompanying drawings, a high-precision map updating device based on point clouds according to an embodiment of this application is described.

[0087] Figure 5 This is a block diagram of a point cloud-based high-precision map updating device according to an embodiment of this application.

[0088] like Figure 5 As shown, the point cloud-based high-precision map update device 10 includes: an acquisition module 100, a processing module 200, and an update module 300.

[0089] The acquisition module 100 is used to acquire point cloud data, pose data and image data of the current location of the vehicle.

[0090] Processing module 200 is used to obtain a local point cloud map based on pose data and a preset point cloud map, and to obtain the optimal matching pose result that meets preset accuracy conditions based on the local point cloud map and point cloud data; and

[0091] The update module 300 is used to update the current local high-precision map based on the optimal matching pose result and image data.

[0092] Furthermore, in some embodiments, the processing module 200 includes:

[0093] The acquisition unit is used to acquire a key frame sequence in a preset point cloud map, wherein the key frame sequence includes the number and pose of each key frame;

[0094] The building unit is used to construct a Kd tree based on the keyframe sequence and determine the keyframe number with the smallest pose distance from the pose data from the Kd tree.

[0095] The stitching unit is used to stitch together a local point cloud map based on the minimum keyframe number in the first preset frame number in the first direction and the minimum keyframe number in the second preset frame number in the second direction.

[0096] Furthermore, in some embodiments, the processing module 200 includes:

[0097] The optimization unit is used to optimize point cloud data based on a local point cloud map using a preset ICP algorithm to obtain the optimal matching pose result and the matching score corresponding to the optimal matching pose result.

[0098] The determination unit is used to determine that the optimal matching pose result meets the preset accuracy condition when the matching score is greater than or equal to the preset score.

[0099] Furthermore, in some embodiments, after obtaining the optimal matching pose result and the matching score corresponding to the optimal matching pose result, the optimization unit is further configured to:

[0100] If the matching score is less than the preset score, the optimal matching pose result is determined to not meet the preset accuracy condition, and the update of the current local high-precision map is stopped.

[0101] Furthermore, in some embodiments, the update module 300 is specifically used for:

[0102] The current frame image is processed using a pre-defined deep learning algorithm to obtain multiple map elements;

[0103] Take any one of the multiple map elements as the current map element, and determine whether there is a target map element in the current local high-precision map that corresponds to the current map element;

[0104] If there is no target map element corresponding to the current map element in the current local high-definition map, the current map element is added to the current local high-definition map. Otherwise, if the matching score of the current map element is greater than the matching score of the target map element, the current map element replaces the target map element. If the matching score of the current map element is less than or equal to the matching score of the target map element, the current local high-definition map is not updated. This process continues until multiple map elements have been traversed.

[0105] It should be noted that the foregoing explanation of the embodiment of the high-precision map update method based on point cloud also applies to the high-precision map update device based on point cloud in this embodiment, and will not be repeated here.

[0106] The high-precision map updating device based on point cloud proposed in this application obtains a local point cloud map by using the pose data of the vehicle's current location and a preset point cloud map. Then, it obtains an optimal matching pose result that meets preset accuracy conditions based on the local point cloud map and the point cloud data of the vehicle's current location. Based on the optimal matching pose result, the current local high-precision map is updated using image data. Therefore, by updating the local high-definition map using point cloud data, pose data, and image data, this solves the problems of error variations caused by multiple vehicles collecting data on the same road segment or multiple collections by a single vehicle in current updating methods, leading to decreased global consistency of the high-precision map and sudden changes in high-precision map elements. This achieves automatic updating of elements in the high-precision map without misalignment, map blurring, or overlay.

[0107] Figure 6 A schematic diagram of the structure of a vehicle provided in an embodiment of this application. The vehicle may include:

[0108] The memory 601, the processor 602, and the computer program stored on the memory 601 and capable of running on the processor 602.

[0109] When the processor 602 executes the program, it implements the point cloud-based high-precision map update method provided in the above embodiments.

[0110] Furthermore, the vehicle also includes:

[0111] Communication interface 603 is used for communication between memory 601 and processor 602.

[0112] The memory 601 is used to store computer programs that can run on the processor 602.

[0113] The memory 601 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0114] If the memory 601, processor 602, and communication interface 603 are implemented independently, then the communication interface 603, memory 601, and processor 602 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0115] Optionally, in a specific implementation, if the memory 601, processor 602, and communication interface 603 are integrated on a single chip, then the memory 601, processor 602, and communication interface 603 can communicate with each other through an internal interface.

[0116] The processor 602 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0117] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described high-precision map update method based on point clouds.

[0118] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0119] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0120] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A high-definition map updating method based on a point cloud, characterized in that, The method comprises the following steps: obtaining point cloud data, pose data and image data of a current position of a vehicle; obtaining a local point cloud map according to the pose data and a preset point cloud map, and obtaining an optimal matching pose result satisfying a preset accuracy condition according to the local point cloud map and the point cloud data; and updating a current local high-precision map according to the image data based on the optimal matching pose result; wherein the obtaining of the local point cloud map according to the pose data and the preset point cloud map comprises: obtaining a key frame sequence in the preset point cloud map, wherein the key frame sequence comprises the number and pose of each key frame; constructing a Kd tree according to the key frame sequence, and determining a key frame number with the minimum distance from the pose data in the Kd tree; and splicing a first preset frame number data of the minimum key frame number in a first direction and a second preset frame number data of the minimum key frame number in a second direction to obtain the local point cloud map; the obtaining of the optimal matching pose result satisfying the preset accuracy condition according to the local point cloud map and the point cloud data comprises: optimizing the point cloud data by using a preset iterative closest point (ICP) algorithm based on the local point cloud map, to obtain the optimal matching pose result and a matching score corresponding to the optimal matching pose result, and determining that the optimal matching pose result satisfies the preset accuracy condition if the matching score is greater than or equal to a preset score.

2. The method of claim 1, wherein, After obtaining the optimal matching pose result and the matching score corresponding to the optimal matching pose result, the method further comprises: determining that the optimal matching pose result does not satisfy the preset accuracy condition if the matching score is less than the preset score, and stopping the updating of the current local high-precision map.

3. The method of claim 1, wherein, the updating of the current local high-precision map according to the image data based on the optimal matching pose result comprises: processing a current frame image by using a preset deep learning algorithm to obtain a plurality of map elements; taking any map element in the plurality of map elements as a current map element, and determining whether a target map element corresponding to the current map element exists in the current local high-precision map; if the target map element corresponding to the current map element does not exist in the current local high-precision map, adding the current map element to the current local high-precision map, otherwise, replacing the target map element with the current map element when a matching score of the current map element is greater than a matching score of the target map element, and not updating the current local high-precision map when the matching score of the current map element is less than or equal to the matching score of the target map element; and 4. A high-definition map updating device based on a point cloud, characterized by, comprising: an obtaining module configured to obtain point cloud data, pose data and image data of a current position of a vehicle; a processing module configured to obtain a local point cloud map according to the pose data and a preset point cloud map, and obtain an optimal matching pose result satisfying a preset accuracy condition according to the local point cloud map and the point cloud data; and and An updating module is configured to update a current local high-definition map according to the image data based on the optimal matching pose result. The processing module is specifically configured to: acquire a key frame sequence in the preset point cloud map, wherein the key frame sequence comprises a number and a pose of each key frame, construct a Kd tree according to the key frame sequence, and determine a key frame number with a minimum distance from a pose in the pose data from the Kd tree, and splice a first preset frame number data in a first direction of the minimum key frame number and a second preset frame number data in a second direction of the minimum key frame number to obtain the local point cloud map. The processing module is specifically configured to: based on the local point cloud map, optimize the point cloud data by using a preset ICP algorithm to obtain the optimal matching pose result and a matching score corresponding to the optimal matching pose result, and if the matching score is greater than or equal to a preset score, determine that the optimal matching pose result satisfies the preset accuracy condition.

5. A vehicle characterized by comprising: The computer program is stored in the memory and executable on the processor, and the processor executes the program to implement the point cloud-based high-definition map updating method according to any one of claims 1-3. The program is executed by the processor to implement the point cloud-based high-definition map updating method according to any one of claims 1-3.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, ​

Images