Vehicle navigation method, device and equipment based on wall surface recognition and storage medium

Abstract

The embodiment of the invention provides a vehicle navigation method and device based on wall surface recognition, equipment and a storage medium. The method comprises the following steps: acquiring radar point cloud data corresponding to a scene where an automatic driving vehicle is located in real time; according to a principal component analysis algorithm, determining a global main direction of the radar point cloud data and a local main direction of each radar point cloud data; performing irregular point elimination processing on the radar point cloud data according to the global main direction and the local main direction to obtain processed point cloud data; performing wall surface identification processing on the processed point cloud data to obtain target wall surface data; determining lane line data of a road where the autonomous vehicle is located according to the target wall data, and determining navigation data of the autonomous vehicle according to the lane line data; wherein the navigation data is used for providing navigation for the autonomous vehicle. According to the method, the accuracy and effectiveness of positioning of the autonomous vehicle can be improved, and stable and reliable navigation is provided for the autonomous vehicle.

Description

Technical Field

[0001] This application relates to the field of automotive technology, and in particular to a vehicle navigation method, apparatus, device, and storage medium based on wall recognition. Background Technology

[0002] Because structured scenarios (such as port scenarios, customs scenarios, etc.) have characteristics such as fixed environment, clear traffic rules, and single task objectives, autonomous vehicles are widely used in structured scenarios.

[0003] However, in structured scenarios, factors such as roof obstruction, metal structure interference, or complex electromagnetic environment may cause autonomous vehicles to face satellite signal failure, leading to positioning failure and navigation failure.

[0004] Furthermore, in structured scenarios, there may be some local irregular features (such as protrusions, pillars, etc.), which further affect the accuracy of autonomous vehicle positioning and thus the navigation performance of autonomous vehicles. Summary of the Invention

[0005] This application provides a vehicle navigation method, apparatus, device, and storage medium based on wall recognition, which can improve the accuracy and effectiveness of autonomous vehicle positioning, thereby providing stable and reliable navigation for autonomous vehicles.

[0006] In a first aspect, embodiments of this application provide a vehicle navigation method based on wall recognition, comprising:

[0007] Real-time acquisition of radar point cloud data corresponding to the scene where the autonomous vehicle is located;

[0008] Based on the principal component analysis algorithm, the global principal direction of the radar point cloud data and the local principal direction of each radar point cloud data are determined.

[0009] Based on the global main direction and the local main direction, the radar point cloud data is subjected to irregular point removal processing to obtain the processed point cloud data;

[0010] The processed point cloud data is subjected to wall recognition processing to obtain target wall data;

[0011] Based on the target wall data, the lane line data of the road where the autonomous vehicle is located is determined, and based on the lane line data, the navigation data of the autonomous vehicle is determined; wherein, the navigation data is used to provide navigation for the autonomous vehicle.

[0012] In one possible implementation, the radar point cloud data is processed by removing irregular points based on the global main direction and the local main direction to obtain processed point cloud data, including:

[0013] Determine the angle data between the local principal direction and the global principal direction;

[0014] If the directional angle data is greater than a preset angle threshold, then the radar point cloud data under the local main direction is determined to be the irregular point;

[0015] The irregular points are removed from the radar point cloud data to obtain the processed point cloud data.

[0016] In one possible implementation, the processed point cloud data is subjected to wall recognition processing to obtain target wall data, including:

[0017] The processed point cloud data is then clustered to obtain at least one initial cluster.

[0018] Based on the at least one initial cluster, determine the wall cluster;

[0019] The wall clusters are subjected to linear fitting to obtain the wall fitting line;

[0020] The target wall data is determined based on the fitted straight line of the wall surface.

[0021] In one possible implementation, determining the wall clusters based on the at least one initial cluster includes:

[0022] Determine the cluster features of each initial cluster in at least one wall feature dimension;

[0023] If it is determined that the cluster features do not meet the preset wall features, then the initial cluster corresponding to the cluster features is removed to obtain the target cluster.

[0024] The wall cluster is determined based on the target cluster.

[0025] In one possible implementation, the number of target clusters is multiple; determining the wall clusters based on the target clusters includes:

[0026] Determine the straight-line distance from each target cluster to the autonomous vehicle;

[0027] If the straight-line distance is not unique, then determine the cluster grouping under each straight-line distance;

[0028] Based on the wall length corresponding to each cluster group, the target cluster is filtered to obtain the wall cluster; wherein, the straight-line distance from the wall cluster to the autonomous vehicle is unique.

[0029] In one possible implementation, the target wall data indicates the wall location; based on the target wall data, determining the lane line data of the road where the autonomous vehicle is located includes:

[0030] The system obtains the first distance between the wall position indicated by the target wall data and the lane line in the road where the autonomous vehicle is located, which is close to the wall position, and the lane width of the road where the autonomous vehicle is located.

[0031] Determine the initial lane line based on the first distance and the lane width;

[0032] The initial lane lines are processed by Kalman filtering to obtain the target lane lines;

[0033] The lane line data is determined based on the target lane line.

[0034] In one possible implementation, before determining the global principal direction of the radar point cloud data according to the principal component analysis algorithm, the method further includes:

[0035] The radar point cloud data is projected to obtain projected radar point cloud data.

[0036] Secondly, embodiments of this application provide a vehicle navigation device based on wall recognition, comprising:

[0037] The acquisition unit is used to acquire radar point cloud data corresponding to the scene where the autonomous vehicle is located in real time.

[0038] The first determining unit is used to determine the global principal direction of the radar point cloud data and the local principal direction of each radar point cloud data according to the principal component analysis algorithm.

[0039] The processing unit is used to perform irregular point removal processing on the radar point cloud data according to the global main direction and the local main direction to obtain the processed point cloud data.

[0040] The recognition unit is used to perform wall recognition processing on the processed point cloud data to obtain target wall data.

[0041] The second determining unit is used to determine the lane line data of the road where the autonomous vehicle is located based on the target wall data, and to determine the navigation data of the autonomous vehicle based on the lane line data; wherein the navigation data is used to provide navigation for the autonomous vehicle.

[0042] In one possible implementation, the processing unit is configured to:

[0043] Determine the angle data between the local principal direction and the global principal direction;

[0044] If the directional angle data is greater than a preset angle threshold, then the radar point cloud data under the local main direction is determined to be the irregular point;

[0045] The irregular points are removed from the radar point cloud data to obtain the processed point cloud data.

[0046] In one possible implementation, the identification unit is used for:

[0047] The processed point cloud data is then clustered to obtain at least one initial cluster.

[0048] Based on the at least one initial cluster, determine the wall cluster;

[0049] The wall clusters are subjected to linear fitting to obtain the wall fitting line;

[0050] The target wall data is determined based on the fitted straight line of the wall surface.

[0051] In one possible implementation, the identification unit is used for:

[0052] Determine the cluster features of each initial cluster in at least one wall feature dimension;

[0053] If it is determined that the cluster features do not meet the preset wall features, then the initial cluster corresponding to the cluster features is removed to obtain the target cluster.

[0054] The wall cluster is determined based on the target cluster.

[0055] In one possible implementation, the number of target clusters is multiple; in this case, the identification unit is used for:

[0056] Determine the straight-line distance from each target cluster to the autonomous vehicle;

[0057] If the straight-line distance is not unique, then determine the cluster grouping under each straight-line distance;

[0058] Based on the wall length corresponding to each cluster group, the target cluster is filtered to obtain the wall cluster; wherein, the straight-line distance from the wall cluster to the autonomous vehicle is unique.

[0059] In one possible implementation, the target wall data indicates the wall location; in this case, the second determining unit is configured to:

[0060] The system obtains the first distance between the wall position indicated by the target wall data and the lane line in the road where the autonomous vehicle is located, which is close to the wall position, and the lane width of the road where the autonomous vehicle is located.

[0061] Determine the initial lane line based on the first distance and the lane width;

[0062] The initial lane lines are processed by Kalman filtering to obtain the target lane lines;

[0063] The lane line data is determined based on the target lane line.

[0064] In one possible implementation, before determining the global principal direction of the radar point cloud data according to the principal component analysis algorithm, the device is further configured to:

[0065] The radar point cloud data is projected to obtain projected radar point cloud data.

[0066] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0067] The memory stores computer-executed instructions;

[0068] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0069] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0070] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0071] The vehicle navigation method, device, equipment, and storage medium based on wall recognition provided in this application can acquire radar point cloud data corresponding to the scene where the autonomous vehicle is located in real time. This allows for the acquisition of richer scene information about the autonomous vehicle through the radar point cloud data. Subsequently, principal component analysis (PCA) can be used to determine the global principal direction and the local principal direction of each radar point cloud data point. This allows for the determination of the local and global data distribution of the radar point cloud data based on the global and local principal directions. Then, irregular points in the radar point cloud data can be filtered based on the global and local principal directions. After removing irregular points, processed point cloud data is obtained. This implementation avoids interference from irregular points, thus preventing point cloud data fitting deviations caused by irregular points, which can lead to positioning and navigation errors. Next, wall recognition processing can be performed on the processed point cloud data to obtain target wall data. Based on the target wall data, lane line data of the road where the autonomous vehicle is located can be determined, and navigation data for the autonomous vehicle can be determined based on the lane line data; the navigation data is used to provide navigation for the autonomous vehicle. In the above embodiment, lane line data can be determined by recognizing the walls in the scene where the autonomous vehicle is located, without relying on high-precision maps, thus reducing costs. At this time, the navigation data for the autonomous vehicle can be determined using the determined lane line data, improving the accuracy and reliability of the navigation data. Attached Figure Description

[0072] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0073] Figure 1 A schematic flowchart illustrating a vehicle navigation method based on wall recognition provided in an embodiment of this application;

[0074] Figure 2 A flowchart illustrating another vehicle navigation method based on wall recognition provided in this application embodiment;

[0075] Figure 3 A scene diagram illustrating the location of an autonomous vehicle, provided as an embodiment of this application;

[0076] Figure 4 A schematic diagram of radar point cloud data corresponding to the scene where an autonomous vehicle is located, provided as an embodiment of this application;

[0077] Figure 5 This is a schematic diagram illustrating an application scenario for determining lane line data, provided in an embodiment of this application.

[0078] Figure 6 A schematic diagram illustrating the implementation process of a vehicle navigation method based on wall recognition provided in this application embodiment;

[0079] Figure 7 A schematic diagram of the structure of a vehicle navigation device based on wall recognition provided in an embodiment of this application;

[0080] Figure 8 This is a schematic diagram of an electronic device structure provided in an embodiment of this application.

[0081] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0082] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0083] Because structured scenarios (such as port scenarios, customs scenarios, etc.) have characteristics such as fixed environment, clear traffic rules, and single task objectives, autonomous vehicles are widely used in structured scenarios.

[0084] However, in structured scenarios, autonomous vehicles may face challenges such as satellite signal failure and poor visual conditions.

[0085] Specifically, in structured environments, factors such as ceiling obstructions, interference from metal structures, or complex electromagnetic environments may cause autonomous vehicles to experience satellite signal failures, leading to positioning failures and ultimately navigation failures. Furthermore, structured environments may contain localized irregularities (such as protrusions or pillars), which can further affect the accuracy of autonomous vehicle positioning and consequently impact navigation performance.

[0086] Furthermore, rainy nights, dirt, wear and tear, dust cover, and insufficient lighting conditions affect the performance of camera-based vehicle positioning methods.

[0087] Research has shown that autonomous vehicles can perceive their environment based on point cloud data collected by LiDAR sensors, and use high-precision point cloud processing algorithms to achieve vehicle positioning and lane line generation. This improves the accuracy and stability of autonomous vehicle positioning, thereby enhancing the precision of autonomous navigation and ensuring the safe and stable operation of autonomous vehicles.

[0088] In one implementation, point cloud data can be fitted using RANSAC (RANdom Sampling Consensus) to extract road boundaries or wall features. However, this implementation is sensitive to local irregularities and is prone to fitting deviations due to these irregularities, resulting in insufficient positioning accuracy.

[0089] In another implementation, point cloud data can be matched with high-precision maps to achieve localization of autonomous vehicles. However, this approach requires the pre-collection of high-precision maps of the scene, resulting in high deployment costs and difficulty in adapting to dynamically changing scenarios. Furthermore, complex algorithms lead to high computational complexity, resulting in high processing latency, which ultimately fails to meet the real-time navigation and control requirements of autonomous vehicles.

[0090] The wall-recognition-based vehicle navigation method provided in this application utilizes principal component analysis (PCA) to remove irregular points from radar point cloud data. Then, the remaining point cloud data, i.e., the processed point cloud data, undergoes wall recognition processing. This enables accurate localization of autonomous vehicles in complex, structured scenes based on the identified target wall data. This implementation method is applicable not only to structured scenes including irregular features but also eliminates the need for high-precision maps, improving the effectiveness and accuracy of autonomous vehicle localization and thus enhancing the navigation performance of autonomous vehicles.

[0091] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0092] Figure 1 A flowchart illustrating a vehicle navigation method based on wall recognition provided in this application is shown below. Figure 1 As shown, the method includes:

[0093] S101. Real-time acquisition of radar point cloud data corresponding to the scene where the autonomous vehicle is located.

[0094] In one example, radar point cloud data can indicate point cloud data collected by a lidar sensor.

[0095] In one example, radar point cloud data can be obtained by reusing the LiDAR sensors installed in an autonomous vehicle to scan the environmental point cloud data of the scene in which the autonomous vehicle is located. For example, LiDAR sensors installed on the left rear and right front of the autonomous vehicle can be reused.

[0096] Alternatively, embodiments of this application can also add LiDAR sensors to the autonomous vehicle according to actual business needs to scan the environmental point cloud data of the scene where the autonomous vehicle is located, thereby obtaining radar point cloud data. For example, LiDAR sensors can be installed at the front left, rear left, front right, and rear right of the autonomous vehicle, so as to scan more comprehensive environmental point cloud data in complex scenes and obtain radar point cloud data. The number of LiDAR sensors installed is not limited here, and the number should be based on actual needs.

[0097] In one example, the scene in which the autonomous vehicle is located may include irregular features, such as protruding walls or pillars. In this case, more detailed scene data can be obtained through radar point cloud data for subsequent localization and navigation of the autonomous vehicle.

[0098] In one example, after acquiring radar point cloud data, the radar point cloud data can be converted to the vehicle coordinate system and then processed in the vehicle coordinate system.

[0099] S102. Based on the principal component analysis algorithm, determine the global principal direction of the radar point cloud data and the local principal direction of each radar point cloud data.

[0100] In one example, principal component analysis (PCA) can be performed on all radar point cloud data to obtain the global principal direction of the radar point cloud data, and PCA can also be performed on the neighboring point cloud data of each radar point cloud data to obtain the local principal direction of each radar point cloud data.

[0101] Optionally, before determining the global main direction, the region of interest can be determined first, and within the region of interest, the global main direction of the radar point cloud data and the local main direction of each radar point cloud data can be determined to reduce the amount of data computation and improve efficiency.

[0102] S103. Based on the global main direction and the local main direction, perform irregular point removal processing on the radar point cloud data to obtain the processed point cloud data.

[0103] In one example, the principal component analysis algorithm can be used to determine the data distribution state of radar point cloud data in the global state, i.e., the global principal direction; and to determine the data distribution state of each radar point cloud data in the local state, i.e., the local principal direction. At this time, the difference between the global principal direction and the local principal direction can be used to determine whether a single radar point cloud data conforms to the data distribution state of the global radar point cloud data, thereby identifying irregular points included in the radar point cloud data.

[0104] Optionally, after removing irregular points from the radar point cloud data, regular point cloud data, i.e., processed point cloud data, can be used to accurately identify positioning references, thereby helping to accurately locate autonomous vehicles and provide precise navigation for them.

[0105] Optionally, the positioning reference in this embodiment can be a wall, as described in the process below.

[0106] S104. Perform wall recognition processing on the processed point cloud data to obtain the target wall data.

[0107] In one example, the processed point cloud data may include point cloud data of one or more walls. In this case, after performing wall recognition processing on the processed point cloud data, one or more walls can be obtained.

[0108] For example, in a structured scenario with single-lane traffic, the identified walls can indicate the walls located on both sides of the autonomous vehicle; in a structured scenario with multi-lane traffic (including one-way and two-way traffic), the identified walls can indicate the wall closer to the autonomous vehicle.

[0109] Optionally, the number of walls located on one side of the autonomous vehicle can be one or more. For example, if the walls on one side of the autonomous vehicle include raised walls and non-raised walls, multiple walls can be identified.

[0110] At this point, the target wall data can be obtained by filtering one or more identified walls.

[0111] Optionally, the target wall data may indicate at least one of the following: wall location, wall length, and wall thickness.

[0112] S105. Based on the target wall data, determine the lane line data of the road where the autonomous vehicle is located, and based on the lane line data, determine the navigation data of the autonomous vehicle.

[0113] The navigation data is used to provide navigation for autonomous vehicles.

[0114] Optionally, after determining the lane line data of the road where the autonomous vehicle is located based on the target wall data, a virtual lane line can be determined based on the lane line data, and the navigation data of the autonomous vehicle can be determined based on the virtual lane line.

[0115] As described above, this embodiment of the application can acquire radar point cloud data corresponding to the scene where the autonomous vehicle is located in real time. This allows for the acquisition of richer information about the scene. Then, based on principal component analysis (PCA) algorithms, the global principal direction and the local principal direction of each radar point cloud data point can be determined. This allows for the determination of the local and global data distribution of the radar point cloud data based on these directions. Next, irregular points in the radar point cloud data can be filtered based on the global and local principal directions. After removing these irregular points, processed point cloud data is obtained. This implementation avoids interference from irregular points, thus preventing point cloud data fitting deviations caused by irregular points, which can lead to positioning and navigation errors. Finally, wall recognition processing can be performed on the processed point cloud data to obtain target wall data. Based on the target wall data, lane line data of the road where the autonomous vehicle is located can be determined, and navigation data for the autonomous vehicle can be determined based on the lane line data. This navigation data is used to provide navigation for the autonomous vehicle. In the above implementation, lane line data can be determined by identifying the walls in the scene where the autonomous vehicle is located, without relying on high-precision maps, thus reducing costs. The determined lane line data can then be used to determine the navigation data for the autonomous vehicle, improving the accuracy and reliability of the navigation data.

[0116] Figure 2 A flowchart illustrating another vehicle navigation method based on wall recognition provided in this application embodiment is shown below. Figure 2 As shown, in this embodiment... Figure 1 Based on the embodiments, the vehicle navigation method based on wall recognition is described in detail. The method includes:

[0117] S201. Real-time acquisition of radar point cloud data corresponding to the scene where the autonomous vehicle is located.

[0118] For example, see Figure 3 , Figure 3 This application provides a schematic diagram of a scenario in which an autonomous vehicle is located, as shown in the embodiments of this application. Figure 3 As shown, the scene in which an autonomous vehicle is located may include irregular features such as protrusions and pillars.

[0119] At this point, environmental point cloud data can be collected by the lidar sensors installed in the autonomous vehicle, and the radar point cloud data corresponding to the scene where the autonomous vehicle is located can be obtained.

[0120] For example, see Figure 4 , Figure 4 This is a schematic diagram of radar point cloud data corresponding to the scene where an autonomous vehicle is located, provided in an embodiment of this application. Figure 4 As shown, irregular features can be accurately identified based on radar point cloud data, for example, Figure 4 The protrusions and columns shown.

[0121] Subsequently, wall recognition and lane line data can be determined based on radar point cloud data, thereby assisting autonomous vehicles in navigation. See the process described below for details.

[0122] S202. Project the radar point cloud data to obtain the projected radar point cloud data.

[0123] In one example, when projecting radar point cloud data, the radar point cloud data can be projected onto the ground, thereby reducing the dimensionality of the three-dimensional data to two-dimensional data, resulting in the projected radar point cloud data.

[0124] At this point, on the one hand, the amount of data computation can be further reduced and the computation efficiency improved; on the other hand, by projecting the wall data onto the base of the wall, the wall data can be made to be more linear, which helps to make accurate wall recognition in the future.

[0125] S203. Based on the principal component analysis algorithm, determine the global principal direction of the radar point cloud data and the local principal direction of each radar point cloud data.

[0126] In one example, this step can be referred to the content described in S102 above, and will not be repeated in detail here.

[0127] S204. Determine the angle data between the local principal direction and the global principal direction.

[0128] S205. If the directional angle data is greater than the preset angle threshold, then the radar point cloud data under the local main direction is determined to be irregular points.

[0129] In one example, the preset angle threshold can be a fixed threshold pre-set according to the scene type of the autonomous vehicle's environment. For example, in a scene with dense point cloud data, the preset angle threshold can be 50 degrees; in a scene with sparse point cloud data, the preset angle threshold can be 60 degrees, etc. The value of the preset angle threshold is not limited here, but is determined by meeting actual needs.

[0130] For example, the radar point cloud data of the corner caused by the protrusion of the wall can have a directional angle between the local main direction and the global main direction of 90 degrees. At this time, the directional angle data is greater than the preset angle threshold. At this time, it can be determined that the radar point cloud data of the corner is an irregular point.

[0131] S206. Remove irregular points from the radar point cloud data to obtain the processed point cloud data.

[0132] For example, irregular points on the wall can be removed by removing the radar point cloud data corresponding to the corner portion mentioned above, that is, the removal of concave and convex parts on the wall.

[0133] This implementation method allows the processed point cloud data to eliminate interference from uneven surfaces and retain the point cloud data of the flat wall surface, thereby helping to improve the accuracy of subsequent wall surface recognition processing.

[0134] Specifically, the process of performing wall recognition processing on the processed point cloud data to obtain the target wall data can be found in the process described below.

[0135] S207. Perform clustering on the processed point cloud data to obtain at least one initial cluster.

[0136] In one possible implementation, before performing clustering on the processed point cloud data, statistical filtering can be performed on the processed point cloud data to remove stray points in the processed point cloud data, thereby ensuring the accuracy and effectiveness of the clustering process.

[0137] Optionally, the processed point cloud data can be clustered according to Euclidean clustering to obtain at least one initial cluster.

[0138] S208. Determine the wall cluster based on at least one initial cluster.

[0139] In practice, feature extraction can be performed on the initial clusters, and at least one initial cluster can be screened based on the extracted features to obtain the wall clusters.

[0140] For example, the cluster features of each initial cluster can be determined first under at least one wall feature dimension; if the cluster features are determined not to meet the preset wall features, the initial clusters corresponding to the cluster features are removed to obtain the target clusters; and the wall clusters are determined based on the target clusters.

[0141] Optionally, the cluster features under the wall feature dimension can indicate, but are not limited to, length feature information and thickness feature information. In this case, the preset wall features can indicate, but are not limited to, preset length information and preset thickness information. Specifically, the information of the initial cluster along the driving direction of the autonomous vehicle is the length feature information; the information of the initial cluster along the left-hand direction of the autonomous vehicle is the thickness feature information.

[0142] At this point, if the length feature information indicates that the wall length is short (for example, the length feature information indicates that the wall length is less than the preset length information, for example, 30 cm), it means that the cluster feature does not meet the preset wall feature; and / or, if the thickness feature information indicates that the wall thickness is thick (for example, the thickness feature information indicates that the wall thickness is greater than the preset thickness information, for example, 50 cm), it means that the cluster feature does not meet the preset wall feature.

[0143] It should be noted that the embodiments of this application do not limit the number and type of cluster features under the wall feature dimension, nor do they limit the specific content of the preset wall features, and shall be based on actual needs.

[0144] This implementation method can delete initial clusters that do not conform to the wall features by presetting the wall features and the cluster features of the initial clusters, thereby further eliminating interference from non-wall clusters (e.g., the initial clusters corresponding to columns).

[0145] In one possible implementation, if there are multiple target clusters determined according to the steps described above, then when determining the wall clusters based on the target clusters, the straight-line distance from each target cluster to the autonomous vehicle can be determined.

[0146] If the straight-line distance is not unique, then the cluster grouping under each straight-line distance is determined, and the target cluster is filtered according to the wall length corresponding to each cluster group to obtain the wall cluster; among them, the straight-line distance from the wall cluster to the autonomous vehicle is unique.

[0147] Optionally, straight-line distance can be understood as the perpendicular distance between the straight line containing the target cluster and the centerline point of the autonomous vehicle.

[0148] If the straight-line distance is not unique, the target cluster is determined to include both convex (or concave) and normal wall portions. Based on the lengths of the convex (or concave) and normal wall portions, the primary wall that the autonomous vehicle relies on during its operation can be identified. For example, the longer wall portion can be identified as the primary wall that the autonomous vehicle relies on during its operation, and a wall cluster can be determined based on this portion.

[0149] For example, if the length of the normal wall portion is 40 meters and the length of the convex wall portion is 10 meters, then the target cluster corresponding to the normal wall portion can be determined as the wall cluster.

[0150] Optionally, if there are multiple wall clusters, the multiple wall clusters can be merged to obtain a merged wall cluster, so as to obtain continuous target wall data.

[0151] In the above implementation, when the wall surface is irregular, the target cluster can be further filtered to obtain wall clusters that conform to the driving scenario of the current autonomous vehicle, thereby improving the accuracy of the target wall data and the accuracy of vehicle navigation.

[0152] S209. Perform linear fitting on the wall clusters to obtain the fitted straight line for the wall.

[0153] S210. Determine the target wall data based on the fitted straight line of the wall surface.

[0154] S211. Based on the target wall data, determine the lane line data of the road where the autonomous vehicle is located, and based on the lane line data, determine the navigation data of the autonomous vehicle.

[0155] The navigation data is used to provide navigation for autonomous vehicles.

[0156] In one possible implementation, the target wall data can indicate the wall location.

[0157] At this point, when determining the lane line data of the road where the autonomous vehicle is located based on the target wall data, the first distance between the wall position indicated by the target wall data and the lane line near the wall position in the road where the autonomous vehicle is located, as well as the lane width of the road where the autonomous vehicle is located, can be obtained. Then, based on the first distance and the lane width, the initial lane line is determined. Next, Kalman filtering is applied to the initial lane line to obtain the target lane line. Finally, the lane line data is determined based on the target lane line.

[0158] In one example, the initial lane lines may include lane lines on both sides of the autonomous vehicle.

[0159] In one example, the initial distance between the wall location and the lane line near the wall location, as well as the lane width of the road where the autonomous vehicle is located, can be prior values.

[0160] For example, see Figure 5 , Figure 5 This application provides an illustration of an application scenario for determining lane line data, as shown in the embodiments of this application. Figure 5As shown, after determining the target wall data, the lane lines closest to the wall can be determined based on the first distance and the target wall data. Then, based on the lane width and the lane lines closest to the wall, another lane line can be determined, thus obtaining the initial lane lines. After performing Kalman filtering on the initial lane lines, the target lane lines are obtained. At this point, the lane line data can be determined based on the target lane lines.

[0161] In the above implementation, accurate lane lines can be determined based on prior knowledge. Furthermore, Kalman filtering can be used to improve the stability and reliability of the output target lane lines, thereby enhancing the accuracy and reliability of the lane line data. In this case, reliable lane line data can provide precise navigation for autonomous vehicles, thus improving their performance.

[0162] See Figure 6 , Figure 6 This application provides an implementation flowchart of a vehicle navigation method based on wall recognition, as shown in the following example. Figure 6 As shown, the vehicle navigation method based on wall recognition specifically includes the following steps:

[0163] The first step is to reuse the LiDAR sensors installed in the autonomous vehicle to obtain radar point cloud data corresponding to the scene where the autonomous vehicle is located.

[0164] The second step is to convert the radar point cloud data to the vehicle coordinate system and set the region of interest to obtain the radar point cloud data of interest. In this vehicle coordinate system, the center point of the autonomous vehicle is the origin, the forward direction of the autonomous vehicle is the positive x-axis, the left side of the autonomous vehicle is the positive y-axis, and the top of the autonomous vehicle is the positive z-axis. At this point, when the vehicle is approximately parallel to the wall, the x-axis represents the extension direction of the wall, and the y-axis represents the left and right directions of the wall.

[0165] The third step is to project the radar point cloud data onto the ground to obtain the projected radar point cloud data. At this point, the point cloud data corresponding to the wall is projected onto the base of the wall, approximating a line, which facilitates straight line fitting.

[0166] The fourth step involves using principal component analysis (PCA) to determine the global principal direction and the local principal direction of each projected radar point cloud data. Based on these directions, irregular points are removed from the projected radar point cloud data to obtain processed point cloud data. At this point, irregular points with large angles between the local and global principal directions can be removed using a preset angle threshold. This process eliminates interference from corners and other concave / convex points, thus separating the corner-connected portions and forming relatively dispersed block / strip-shaped point cloud data.

[0167] Step 5: Perform statistical filtering on the processed point cloud data to eliminate stray points in the processed point cloud data and obtain filtered point cloud data.

[0168] Step 6: Perform clustering processing on the filtered point cloud data to obtain at least one initial cluster, and remove the initial clusters that do not meet the preset wall features to obtain the target cluster.

[0169] Step 7: If there is only one target cluster, then that target cluster is designated as the wall cluster. If there are multiple target clusters, then the straight-line distance from each target cluster to the autonomous vehicle can be determined. In this case, if the straight-line distance is unique, the target clusters are merged to obtain the wall cluster; if the straight-line distance is not unique, then the cluster groups under each straight-line distance are determined. Then, based on the wall length corresponding to each cluster group, the target clusters are filtered to obtain the wall cluster.

[0170] Then, a straight line fitting process can be performed on the wall cluster to obtain the wall fitting line, which is the target wall data.

[0171] Step 8: Determine the initial lane lines based on the target wall data, the first distance between the wall position indicated by the target wall data and the lane line near the wall position in the road where the autonomous vehicle is located, and the lane width of the road where the autonomous vehicle is located.

[0172] Step 9: Perform Kalman filtering on the initial lane lines to obtain the target lane lines, and determine the lane line data based on the target lane lines.

[0173] Step 10: Based on the lane line data, determine the navigation data for the autonomous vehicle, and provide navigation for the autonomous vehicle based on the navigation data.

[0174] Figure 7 A schematic diagram of a vehicle navigation device based on wall recognition provided in this application embodiment is shown below. Figure 7 As shown, the wall-based vehicle navigation device 70 provided in this embodiment includes:

[0175] The acquisition unit 701 is used to acquire radar point cloud data corresponding to the scene where the autonomous vehicle is located in real time.

[0176] The first determining unit 702 is used to determine the global principal direction of radar point cloud data and the local principal direction of each radar point cloud data according to the principal component analysis algorithm.

[0177] The processing unit 703 is used to perform irregular point removal processing on the radar point cloud data according to the global main direction and the local main direction to obtain the processed point cloud data.

[0178] The recognition unit 704 is used to perform wall recognition processing on the processed point cloud data to obtain target wall data.

[0179] The second determining unit 705 is used to determine the lane line data of the road where the autonomous vehicle is located based on the target wall data, and to determine the navigation data of the autonomous vehicle based on the lane line data; wherein the navigation data is used to provide navigation for the autonomous vehicle.

[0180] In one possible implementation, the processing unit 703 is configured to:

[0181] Determine the angle between the local principal direction and the global principal direction;

[0182] If the directional angle data is greater than the preset angle threshold, then the radar point cloud data under the local main direction is determined to be irregular points;

[0183] Irregular points are removed from the radar point cloud data to obtain the processed point cloud data.

[0184] In one possible implementation, the identification unit 704 is used for:

[0185] The processed point cloud data is clustered to obtain at least one initial cluster.

[0186] Determine the wall clusters based on at least one initial cluster;

[0187] The wall clusters are subjected to linear fitting to obtain the fitted straight line for the wall.

[0188] The target wall data is determined by fitting a straight line to the wall surface.

[0189] In one possible implementation, the identification unit 704 is used for:

[0190] Determine the cluster features of each initial cluster in at least one wall feature dimension;

[0191] If it is determined that the cluster features do not meet the preset wall features, the initial cluster corresponding to the cluster features is removed to obtain the target cluster.

[0192] Based on the target cluster, determine the wall cluster.

[0193] In one possible implementation, the number of target clusters is multiple; in this case, the identification unit 704 is used for:

[0194] Determine the straight-line distance from each target cluster to the autonomous vehicle;

[0195] If the straight-line distance is not unique, then determine the cluster grouping for each straight-line distance;

[0196] Based on the wall length corresponding to each cluster group, the target clusters are filtered to obtain wall clusters; among them, the straight-line distance from the wall cluster to the autonomous vehicle is unique.

[0197] In one possible implementation, the target wall data indicates the wall location; at this time, the second determining unit 705 is used to:

[0198] The first distance between the wall position indicated by the target wall data and the lane line near the wall position in the road where the autonomous vehicle is located, as well as the lane width of the road where the autonomous vehicle is located;

[0199] Determine the initial lane lines based on the first distance and lane width;

[0200] The initial lane lines are processed by Kalman filtering to obtain the target lane lines;

[0201] Determine lane line data based on the target lane lines.

[0202] In one possible implementation, before determining the global principal direction of the radar point cloud data according to the principal component analysis algorithm, the device is further used to:

[0203] The radar point cloud data is projected to obtain the projected radar point cloud data.

[0204] The vehicle navigation device based on wall recognition provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0205] Figure 8 This is a schematic diagram of an electronic device structure provided as an embodiment of this application. Figure 8 As shown, the electronic device 80 provided in this embodiment includes at least one processor 801 and a memory 802. Optionally, the electronic device 80 further includes a communication component 803. The processor 801, memory 802, and communication component 803 are connected via a bus 804.

[0206] In a specific implementation, at least one processor 801 executes computer execution instructions stored in memory 802, causing at least one processor 801 to perform the above-described method.

[0207] The specific implementation process of processor 801 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0208] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0209] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0210] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0211] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0212] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0213] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0214] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0215] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0216] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0217] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0218] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0219] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0220] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A vehicle navigation method based on wall recognition, characterized in that, include: Real-time acquisition of radar point cloud data corresponding to the scene where the autonomous vehicle is located; Based on the principal component analysis algorithm, the global principal direction of the radar point cloud data and the local principal direction of each radar point cloud data are determined. Based on the global main direction and the local main direction, the radar point cloud data is subjected to irregular point removal processing to obtain the processed point cloud data; The processed point cloud data is subjected to wall recognition processing to obtain target wall data; Based on the target wall data, the lane line data of the road where the autonomous vehicle is located is determined, and based on the lane line data, the navigation data of the autonomous vehicle is determined; wherein, the navigation data is used to provide navigation for the autonomous vehicle.

2. The method according to claim 1, characterized in that, Based on the global principal direction and the local principal direction, irregular point removal processing is performed on the radar point cloud data to obtain processed point cloud data, including: Determine the angle data between the local principal direction and the global principal direction; If the directional angle data is greater than a preset angle threshold, then the radar point cloud data under the local main direction is determined to be the irregular point; The irregular points are removed from the radar point cloud data to obtain the processed point cloud data.

3. The method according to claim 1, characterized in that, The processed point cloud data is subjected to wall recognition processing to obtain target wall data, including: The processed point cloud data is then clustered to obtain at least one initial cluster. Based on the at least one initial cluster, determine the wall cluster; The wall clusters are subjected to linear fitting to obtain the wall fitting line; The target wall data is determined based on the fitted straight line of the wall surface.

4. The method according to claim 3, characterized in that, Determining wall clusters based on the at least one initial cluster includes: Determine the cluster features of each initial cluster in at least one wall feature dimension; If it is determined that the cluster features do not meet the preset wall features, then the initial cluster corresponding to the cluster features is removed to obtain the target cluster. The wall cluster is determined based on the target cluster.

5. The method according to claim 4, characterized in that, The number of target clusters is multiple; the wall clusters are determined based on the target clusters, including: Determine the straight-line distance from each target cluster to the autonomous vehicle; If the straight-line distance is not unique, then determine the cluster grouping under each straight-line distance; Based on the wall length corresponding to each cluster group, the target cluster is filtered to obtain the wall cluster; wherein, the straight-line distance from the wall cluster to the autonomous vehicle is unique.

6. The method according to any one of claims 1-5, characterized in that, The target wall data indicates the wall's location; based on the target wall data, the lane markings of the road where the autonomous vehicle is located are determined, including: The system obtains the first distance between the wall position indicated by the target wall data and the lane line in the road where the autonomous vehicle is located, which is close to the wall position, and the lane width of the road where the autonomous vehicle is located. Determine the initial lane line based on the first distance and the lane width; The initial lane lines are processed by Kalman filtering to obtain the target lane lines; The lane line data is determined based on the target lane line.

7. The method according to any one of claims 1-5, characterized in that, Before determining the global principal direction of the radar point cloud data based on the principal component analysis algorithm, the method further includes: The radar point cloud data is projected to obtain projected radar point cloud data.

8. A vehicle navigation device based on wall recognition, characterized in that, include: The acquisition unit is used to acquire radar point cloud data corresponding to the scene where the autonomous vehicle is located in real time. The first determining unit is used to determine the global principal direction of the radar point cloud data and the local principal direction of each radar point cloud data according to the principal component analysis algorithm. The processing unit is used to perform irregular point removal processing on the radar point cloud data according to the global main direction and the local main direction to obtain the processed point cloud data. The recognition unit is used to perform wall recognition processing on the processed point cloud data to obtain target wall data. The second determining unit is used to determine the lane line data of the road where the autonomous vehicle is located based on the target wall data, and to determine the navigation data of the autonomous vehicle based on the lane line data; wherein the navigation data is used to provide navigation for the autonomous vehicle.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images