Point cloud data processing method, device and equipment, and readable storage medium

Abstract

The embodiment of the invention provides a point cloud data processing method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining a plurality of groups of scanning data which are laser point cloud data of an intersection environment collected by a vehicle in a plurality of scanning times; determining a predicted scanning track and position information of scanning times corresponding to each group of scanning data; at least part of the scanning data is screened based on a maximum clique algorithm according to the correlation of the multiple predicted scanning tracks in the geographic and / or time dimension, multiple maximum cliques are obtained, and each maximum clique comprises multiple sets of target scanning data; target features in each group of target scanning data are tracked, pose information corresponding to multiple groups of target scanning data in each scanning data set is determined, and the target features are high-reflectivity markers in the intersection environment; and according to the pose information corresponding to each group of target scanning data, splicing the plurality of groups of target scanning data in the plurality of maximum clusters to obtain a three-dimensional map of the intersection environment.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and specifically to a point cloud data processing method, device and equipment, and a readable storage medium. Background Art

[0002] Light Laser Detection and Ranging (LiDAR) systems are increasingly being used in geographic information systems and autonomous vehicles due to their high accuracy and ability to work in a variety of lighting and weather conditions.

[0003] When a vehicle equipped with a laser radar passes through an intersection, the laser radar collects point cloud data of the surrounding environment based on the principle of laser reflection and uploads it to the cloud server. The cloud server receives point cloud data collected by different vehicles at different times, performs trajectory alignment, inter-trip loop detection, and pose optimization, and then splices the point cloud data into a seamless high-precision 3D map.

[0004] In the above method, the quality and relevance of the captured data are not considered, but all the captured data are aggregated, which increases the complexity of data processing and significantly increases the consumption of computing resources. At the same time, the processing of multi-pass scanning data relies on the LOAM algorithm of surface and line features, which can show good real-time positioning and map building capabilities in a single scan, but when multiple scans involve different environmental conditions or different time periods, traditional surface and line features may lose stability and significance due to environmental changes. The instability of this feature makes it difficult to achieve high-precision alignment of inter-pass data fusion, which in turn affects the quality and reliability of the overall map. Summary of the invention

[0005] The present application provides a point cloud data processing method, device and equipment, and a readable storage medium. The following introduces various aspects involved in the embodiments of the present application.

[0006] In a first aspect, a point cloud data processing method is provided, the method comprising: acquiring multiple groups of scanning data, the multiple groups of scanning data being laser point cloud data of an intersection environment collected by a vehicle in multiple scanning passes; determining a predicted scanning trajectory of a scanning pass corresponding to each group of scanning data and location information of each scanning pass; screening at least part of the multiple groups of scanning data based on a maximum clique algorithm according to the correlation of the multiple predicted scanning trajectories in geographic and / or temporal dimensions to obtain multiple maximum cliques, wherein each maximum clique includes multiple groups of target scanning data; tracking target features in each group of target scanning data, and determining posture information corresponding to the multiple groups of target scanning data in each scanning data set, wherein the target features are high-reflectivity markers in the intersection environment; and splicing the multiple groups of target scanning data in multiple maximum cliques according to the posture information corresponding to each group of target scanning data to obtain a three-dimensional map of the intersection environment.

[0007] According to the above technical means, when synthesizing a 3D map using point cloud data from multiple passes, multiple sets of scan data are screened by analyzing the correlation of vehicle trajectories to reduce redundant data; when performing feature matching, high-reflectivity markers in the intersection environment are used to achieve stable cross-pass feature matching. This technical solution can improve the accuracy of the 3D map while reducing the amount of calculation.

[0008] In some embodiments, the predicted scanning trajectory of the scanning pass corresponding to each group of scanning data and the position information of each scanning pass are determined, including: using a laser inertial odometer (LIO) algorithm to process each group of scanning data to generate a predicted scanning trajectory for each scanning pass; and using the coordinate information of the starting position of each scanning pass determined by a satellite positioning system as the position information.

[0009] According to the above technical means, the vehicle trajectory and location information of each scanning trip are determined based on the LIO algorithm and the satellite positioning system, which can improve the accuracy of the vehicle scanning trajectory prediction and provide a geographical reference for subsequent processing.

[0010] In some embodiments, before filtering multiple groups of scanning data based on the maximum clique algorithm according to the correlation of multiple predicted trajectories in the geographical and / or temporal dimensions, the method also includes: determining a first predicted driving trajectory of the vehicle in each scanning trip based on the vehicle's inertial measurement unit; determining a second predicted driving trajectory of the vehicle in each scanning trip based on a satellite positioning system; filtering multiple groups of scanning data based on the trajectory correlation of the first predicted driving trajectory, the second predicted driving trajectory, and the predicted scanning trajectory, and eliminating scanning data whose trajectory correlation is less than a first correlation threshold.

[0011] According to the above-mentioned technical means, by performing quality analysis on each single trip data, low-quality data is eliminated based on factors such as data integrity, accuracy and environmental variables, ensuring that the subsequently processed data is reliable and useful, avoiding interference in the subsequent map stitching process, and improving the quality and reliability of the map.

[0012] In some embodiments, at least part of the scanning data in multiple groups of scanning data are screened based on the maximum clique algorithm, including: constructing an undirected graph based on the correlation of multiple predicted scanning trajectories corresponding to at least part of the scanning data in geographical and / or time dimensions; wherein each node in the undirected graph represents a predicted scanning trajectory, and the edge in the undirected graph represents that the correlation of the predicted scanning trajectories corresponding to two nodes in geographical and / or time dimensions is greater than a second correlation threshold; based on the maximum clique algorithm, determining multiple maximum cliques in the undirected graph, wherein each maximum clique includes multiple groups of target scanning data.

[0013] According to the above technical means, based on the correlation of the predicted trajectories of the scanned data, the maximum clique algorithm in graph theory is used to identify and aggregate single-trip data with high correlation. In this process, each identified maximum clique contains trajectories that are closely related to each other. The maximum clique found represents the set of trajectories that should be aggregated the most. Such aggregation can significantly reduce the need to process redundant data, thereby improving the overall data processing efficiency and system performance.

[0014] In some embodiments, target features in each set of target scanning data are tracked, and posture information corresponding to multiple sets of target scanning data in each scanning data set is determined, including: tracking multiple feature points corresponding to the target features in each set of target scanning data, and determining the translation and rotation relationship of each set of target scanning data relative to the target features.

[0015] According to the above technical means, by identifying the target features of the intersection environment, the extracted features are used to construct a detailed local map for each trip to provide sufficient details to support accurate matching.

[0016] In some embodiments, tracking the target features in each set of target scanning data and determining the position information corresponding to multiple sets of target scanning data in each scanning data set also includes: tracking multiple feature points corresponding to the target features in each set of target scanning data, and determining the position deviation and angle deviation between different target data based on the different shapes, sizes and angles of the target features in different target scanning data; wherein the target features are road signs in an intersection environment.

[0017] According to the above technical means, highly reflective road signs are used as target features, and the strong reflection signals emitted by them are captured by the laser scanner and can be identified as stable features. By identifying the same feature points in different passes of data, it is ensured that multiple passes of data can be correctly aligned in space, overcoming the instability problem of traditional methods based on surface and line features in multiple passes of scanning.

[0018] In some embodiments, multiple groups of target scanning data in multiple maximum clusters are spliced ​​according to the posture information corresponding to each group of target scanning data, including: constructing a posture graph according to the posture information corresponding to each group of target scanning data; adjusting the posture of each target scanning data based on a posture graph optimization algorithm; and splicing a three-dimensional map of the intersection environment according to the adjusted posture.

[0019] According to the above technical means, the posture of each single trip is accurately adjusted based on graph optimization technology, and then spliced ​​into a continuous and seamless point cloud map according to the optimized posture, which can ensure the overall consistency and accuracy of the map.

[0020] In a second aspect, a point cloud data processing device is provided, comprising: an acquisition unit, used to acquire multiple groups of scanning data, where the multiple groups of scanning data are laser point cloud data of an intersection environment collected by a vehicle in multiple scanning passes; a first determination unit, used to determine the predicted scanning trajectory of the scanning pass corresponding to each group of scanning data and the position information of each scanning pass; a screening unit, used to screen at least part of the scanning data in the multiple groups of scanning data based on a maximum clique algorithm according to the correlation of the multiple predicted scanning trajectories in the geographical and / or temporal dimensions, so as to obtain multiple maximum cliques; wherein each maximum clique includes multiple groups of target scanning data; a second determination unit, used to track target features in each group of target scanning data, and determine the corresponding posture information of the multiple groups of target scanning data in each scanning data set, wherein the target features are high-reflectivity markers in the intersection environment; a splicing unit, used to splice the multiple groups of target scanning data in multiple maximum cliques according to the posture information corresponding to each group of target scanning data, so as to obtain a three-dimensional map of the intersection environment.

[0021] In a third aspect, a point cloud data processing device is provided, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the method described in the first aspect.

[0022] According to a fourth aspect, a computer-readable storage medium is provided for storing a computer program, and when the computer program is executed, the method described in the first aspect is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic flow chart of a method for generating a three-dimensional map of an intersection in the related art;

[0024] Figure 2 A schematic flow chart of a point cloud data processing method provided in an embodiment of the present application;

[0025] Figure 3 A schematic flow chart of a point cloud data processing method provided in another embodiment of the present application;

[0026] Figure 4 A schematic flow chart of a method for screening scan data in a point cloud data processing method provided in an embodiment of the present application;

[0027] Figure 5 A schematic flow chart of a method for splicing target scan data in a point cloud data processing method provided in an embodiment of the present application;

[0028] Figure 6 A schematic structural diagram of a point cloud data processing device provided in an embodiment of the present application;

[0029] Figure 7 A schematic structural diagram of a point cloud data processing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will describe the implementation methods of the present application with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be understood that the preferred embodiments are only for illustrating the present application, not for limiting the scope of protection of the present application.

[0031] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application, and thus the drawings only show components related to the present application rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed at will, and the component layout may also be more complicated.

[0032] Before introducing the technical solution of the embodiment of the present application, the point cloud data processing method in the related art and the existing problems thereof are firstly illustrated in detail with examples.

[0033] With the continuous development of road traffic, the number of traffic participants such as vehicles, pedestrians, and bicycles on the road is also increasing. Road traffic contains many complex structures (such as roundabouts, overpasses, etc.). For intelligent assisted driving systems, it is a challenge to accurately obtain the geographical location of road traffic, road signs, traffic lights and other traffic objects. The safety of intelligent driving systems in road traffic is crucial. Not only does the system itself need to have high stability and robustness to cope with various complex road conditions, but it also requires the intelligent driving system to understand and comply with the rules and information in road traffic (such as traffic lights, traffic signs, etc.) to ensure the safety of users.

[0034] In highways, urban expressways, bridges and other environments, since the road surface is well marked with signs or markings, the vehicle's intelligent driving system can use these signs or markings as a reference, accurately perceive the surrounding vehicles and obstacles, and perform intelligent assisted driving for the vehicle. However, on urban roads, such environments usually have complex traffic flows and diverse road signs, which puts higher requirements on the perception and decision-making capabilities of intelligent assisted driving. The vehicle needs to be able to recognize and understand various traffic participants such as traffic lights, pedestrians and vehicles, and make corresponding decisions based on real-time traffic conditions.

[0035] In the related technologies, one feasible way to improve the intelligent assisted driving capability of vehicles is to perform assisted driving based on high-precision maps, which first requires the construction of a three-dimensional high-precision map of the road. In the field of map construction, a variety of technologies are widely used, including satellite positioning systems (such as the Global Positioning System, the Galileo Positioning System, and the Beidou Navigation Positioning System), radar detection, and optical camera systems. These technologies support the generation of a variety of maps from simple road maps to complex city models.

[0036] However, despite the widespread application of the above methods, there are still many challenges in constructing high-precision maps in complex environments such as intersections. Especially in multi-lane, traffic-intensive urban environments, the above methods often have difficulty in capturing complete and accurate road surface information.

[0037] Existing map building methods mainly rely on global positioning systems and camera systems, which can provide reasonable accuracy under good environmental conditions, but their effectiveness is greatly reduced in poor visibility or bad weather conditions. In addition, poor coverage and signal reflection problems of positioning signals in urban canyons further limit their application in complex intersection mapping.

[0038] In recent years, Light Laser Detection and Ranging (LiDAR) systems have been increasingly used in geographic information systems and autonomous vehicles due to their high accuracy and ability to work under various lighting and weather conditions.

[0039] Laser scanning does not rely on external light sources. It generates accurate three-dimensional coordinates of objects based on the principle of laser reflection, which is particularly important for mapping complex intersections. Figure 1 The figure shows a schematic flow chart of a method for generating a three-dimensional map of an intersection based on laser scanning technology in the related art. The method includes steps S110-S140.

[0040] Step S110, rough alignment of trajectories. Use GPS data to perform rough alignment of trajectories. The trajectory referred to here is the movement trajectory of a vehicle or other acquisition device when using a laser radar to acquire a scanned data set. Each laser scanned data set is associated with a GPS coordinate, which is used to initially convert the data set into a unified coordinate system. The GPS coordinate is usually the coordinate of the starting position of the trajectory.

[0041] Step S120, loop detection between trips. The loop detection compares the features in each scan data set with the recorded feature library to determine whether the same location is repeated. The recorded feature library can be fixed objects in the intersection environment, such as curbs, fixed traffic lights at intersections, etc.

[0042] Step S130, pose optimization. According to the results of loop detection, a pose graph is constructed. The nodes represent the positions of each scan, and the edges represent the relative poses between the passes. The pose graph is optimized to accurately correct the position and direction deviations between the passes.

[0043] Step S140: point cloud stitching: Based on the optimized posture data, the point cloud data of each pass is adjusted and stitched according to the posture to construct a continuous and seamless high-precision three-dimensional map.

[0044] The above method can improve the overall accuracy of the map and enhance the continuity and completeness of the map in details, but it still has the following problems:

[0045] The above method does not consider the quality and relevance of the captured data, but aggregates all the captured data. This method increases the complexity of data processing and significantly increases the consumption of computing resources. At the same time, the processing of multi-pass scanning data relies on the LOAM algorithm of surface and line features, which can show good real-time positioning and map building capabilities in a single scan. However, when multiple scans involve different environmental conditions or different time periods, traditional surface and line features may lose stability and significance due to environmental changes. The instability of this feature makes it difficult to achieve high-precision alignment of inter-pass data fusion, which in turn affects the quality and reliability of the overall map. Therefore, how to ensure fusion accuracy while improving data processing efficiency has become an urgent problem to be solved.

[0046] In view of the above problems, the embodiments of the present application provide a point cloud data processing method, device and equipment, and a readable storage medium. The technical solution provided by the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings.

[0047] Figure 2 It is a schematic flow chart of the point cloud data processing method provided in an embodiment of the present application. The method can be executed by a host computer or a server. The server can be a server deployed locally or a server deployed in the cloud.

[0048] Figure 2 The method includes steps S210-S250.

[0049] In step S210, multiple sets of scanning data are acquired, which are laser point cloud data of the intersection environment collected by the vehicle in multiple scanning passes.

[0050] Laser point cloud data is a data set of spatial points obtained by scanning with a laser radar installed on the vehicle body while the vehicle is moving. The laser radar emits a laser signal, then collects the laser signal reflected by objects in the vehicle's surrounding environment, and combines information such as the speed of light and the time from the laser signal's emission to its return to measure the distance to the target object. It then calculates the three-dimensional coordinate information of each spatial point by combining other information.

[0051] It is understandable that the multiple sets of scanning data may be laser point cloud data obtained by scanning the same vehicle in different scanning passes, or may be collected by different vehicles, and the embodiments of the present application do not specifically limit this.

[0052] It is also understandable that the performance parameters of the laser radars configured for different vehicles are usually different. For example, the number of scanning laser lines and / or scanning frequency are different, which will result in different accuracies of different scanning data and the range of intersection environments covered by the scanning data, which is not limited in the present application.

[0053] There are many ways to obtain the multiple sets of scanning data. For example, the vehicle can automatically send the scanned laser point cloud data to the server in real time through a wireless network during driving; or, the vehicle can store the scanned laser point cloud data in the vehicle's storage unit during driving, and the vehicle operator can actively upload it to the server.

[0054] In step S220, the predicted scanning trajectory of the scanning pass corresponding to each group of scanning data and the position information of each scanning pass are determined.

[0055] It is understood that the coordinates of the point cloud data collected by the vehicle's LiDAR are in a coordinate system with the LiDAR itself as the origin, which can be a Cartesian coordinate system or a polar coordinate system. In order to facilitate the fusion of different data, different point cloud data need to be converted to the same coordinate system. The goal of this step is to prepare and optimize the data of each pass for more complex processing and analysis.

[0056] There can be many methods to determine the predicted scanning trajectory of each scanning pass. For example, the vehicle's inertial measurement unit can be used to predict the scanning trajectory, that is, the vehicle's driving trajectory is predicted by the vehicle's speed and acceleration in all directions during driving; or, it can be based on a satellite positioning system, that is, during the vehicle's movement, the satellite positioning system is used to determine the vehicle's latitude and longitude, and then determine the vehicle's trajectory. The satellite positioning system mentioned here may include the Beidou Satellite Navigation System BDS, the Global Positioning System GPS, the GLONASS Satellite Navigation System GLONASS and the Galileo Satellite Navigation System GALILEO, etc.

[0057] In some embodiments of the present application, the predicted scanning trajectory of each scanning pass can be performed based on a laser inertial odometer (LIO) algorithm, which can use the original point cloud data to generate a high-precision predicted scanning trajectory for a single pass.

[0058] In some embodiments, the scanning trajectory may be determined by combining the prediction results of the aforementioned LIO algorithm, the prediction results of the vehicle inertial measurement unit, and the measurement data of the satellite positioning system, which can improve the accuracy and robustness of the position estimation.

[0059] The location information of each scanning pass may be the initial location of the vehicle at each scanning pass, which may provide a geographical reference.

[0060] In some embodiments, the initial position of each scanning pass can be determined using a satellite positioning system, that is, the vehicle is positioned using the aforementioned satellite positioning systems to determine the longitude and latitude of the initial position.

[0061] As a possible implementation method, in addition to the satellite positioning system, other methods can be used to determine the initial position of the scanning pass. For example, a wireless communication base station can be used to assist in positioning the vehicle, or a real-time dynamic base station RTK positioning method can be used, which has higher dynamic accuracy and can improve the positioning accuracy of the vehicle.

[0062] According to the above technical means, the vehicle trajectory and location information of each scanning trip are determined based on the LIO algorithm and the satellite positioning system, which can improve the accuracy of the vehicle scanning trajectory prediction and provide a geographical reference for subsequent processing.

[0063] In step S230 , at least part of the scan data in the plurality of groups of scan data is screened based on a maximum clique algorithm according to the correlation of the plurality of predicted scan trajectories in the geographical and / or temporal dimensions.

[0064] In an embodiment of the present application, the correlation of the geographic dimension can be determined based on the deviation between the predicted trajectories corresponding to different scanning data or the deviation between the initial positions of the corresponding scanning passes. The larger the deviation, the lower the correlation.

[0065] The maximum clique algorithm is a fast search algorithm, which is used to find the clique with the largest number of vertices from an undirected graph composed of multiple points. Each of the multiple maximum cliques based on the above-mentioned maximum clique algorithm includes multiple groups of target scanning data, and the target scanning data in each maximum clique are scanning data with high correlation in geographical and / or time dimensions. For example, the multiple groups of target scanning data can be multiple groups of point cloud data obtained by scanning along the same or almost the same scanning trajectory. It can be understood that the number of target scanning data in different maximum cliques can be the same or different, and the embodiments of the present application are not limited to this.

[0066] The purpose of this step is to optimize the processing and aggregation of data, improve data quality and processing efficiency, and determine which trajectory sets can be optimally aggregated through the maximum clique algorithm. In this process, each identified maximum clique contains closely related trajectories that overlap significantly geographically or temporally. Such aggregation can significantly reduce the need to process redundant data, thereby improving overall data processing efficiency and system performance.

[0067] In step S240, target features in each set of target data are tracked to determine the position and posture information corresponding to the multiple sets of target data in each scan data set.

[0068] Among them, the target feature is a high-reflectivity marker in the intersection environment. The marker can be, for example, a road sign. In order to enhance the nighttime visibility of the road sign, its surface is usually provided with retro-reflective materials such as reflective film, which can reflect the light of the vehicle for the driver to observe. This type of material is not only highly reflective to visible light, but also has high reflectivity to the invisible light emitted by the laser radar. The aforementioned road signs and other markers will appear as high-density areas in the laser point cloud. Moreover, such markers are usually fixed relative to the intersection and will not change with changes in environmental factors. Therefore, in the technical solution provided in the embodiment of the present application, the position and posture information of each group of target data can be determined by tracking the markers in the point cloud data.

[0069] According to the above technical means, highly reflective road signs are used as target features, and the strong reflection signals emitted by them are captured by the laser scanner and can be identified as stable features, avoiding the instability of traditional methods based on surface and line features.

[0070] Based on the extracted features, a detailed local map is constructed for each pass, providing enough details to support accurate matching. Matching operations are performed between local maps of different passes to identify translation and rotation relationships of the same locations, thereby ensuring that multiple passes of data can be spatially aligned.

[0071] In step S250, multiple groups of target scanning data in multiple maximum clusters are spliced ​​according to the position and posture information corresponding to each group of target scanning data to obtain a three-dimensional map of the intersection environment.

[0072] According to the aforementioned steps, the position and posture information of multiple groups of target data are determined by extracting features. Based on the position and posture information of each group of target data, a continuous and seamless high-precision three-dimensional map can be constructed for each pass of data.

[0073] According to the above method provided by the embodiment of the present application, when synthesizing a three-dimensional map using point cloud data from multiple passes, multiple sets of scan data are screened by analyzing the correlation of vehicle trajectories to reduce redundant data; when performing feature matching, high-reflectivity markers in the intersection environment are used to achieve stable cross-pass feature matching. This technical solution can improve the accuracy of the three-dimensional map while reducing the amount of calculation.

[0074] In some embodiments, Figure 3 As shown, in step S230, according to the correlation of multiple predicted scanning trajectories in the geographical and / or temporal dimensions, before screening at least part of the scanning data in the multiple groups of scanning data based on the maximum clique algorithm, the data processing method provided in the embodiment of the present application also includes steps S311-S313.

[0075] In step S311 , a first predicted driving trajectory of the vehicle in each scanning pass is determined based on an inertial measurement unit of the vehicle.

[0076] An inertial measurement unit typically includes an accelerometer and a gyroscope. The accelerometer is used to measure the acceleration of the vehicle in different directions in three-dimensional space, and the gyroscope is used to measure the rotation speed of the vehicle, helping to determine its angular velocity, and then calculating the vehicle's attitude and heading. In some implementations, the inertial measurement unit may also include a magnetometer, which is used to detect the earth's magnetic field and assist in correcting the drift of the gyroscope. Using the inertial measurement unit, the vehicle's driving trajectory can be estimated based on the vehicle's own motion parameters, that is, the aforementioned first predicted driving trajectory.

[0077] In step S312, a second predicted driving trajectory of the vehicle in each scanning pass is determined based on the satellite positioning system.

[0078] The satellite positioning system may be any one of the multiple satellite positioning systems described above. During the movement of the vehicle, the latitude and longitude information of the vehicle is collected in real time to obtain the second predicted driving trajectory.

[0079] In step S313, based on the correlation among the first predicted driving trajectory, the second predicted driving trajectory and the predicted scanning trajectory, the plurality of groups of scanning data are screened, and scanning data whose trajectory correlation is less than a first correlation threshold is eliminated.

[0080] The correlation of the above multiple trajectories is used to characterize the degree of deviation between different trajectories. For example, the correlation can be evaluated based on the deviation of the initial position, intermediate position or end position of each trajectory; or, each trajectory can be sampled according to the driving time of the vehicle, and multiple positions on different trajectories at the same time are selected, and the correlation is evaluated based on the deviation between the positions corresponding to the above three trajectories at the same time.

[0081] When the correlation between the above multiple trajectories is less than the first correlation threshold, it can be considered that there is a large deviation in the group of data, for example, there is an error in the data collected by the inertial measurement unit, or there are too many interference factors in the intersection environment, resulting in the predicted scanning trajectory disc being the actual situation. In this case, the group of data should be removed to avoid interference in the subsequent splicing process and affect the quality and reliability of the overall map.

[0082] It should be noted that the evaluation index of the correlation of the above-mentioned trajectories and the selection criteria of the first correlation threshold are only examples. In practical applications, appropriate evaluation indexes and specific values ​​of the first correlation threshold can be selected according to the requirements for data processing speed and accuracy. The embodiments of the present application do not make specific limitations on this.

[0083] According to the above-mentioned technical means, by performing quality analysis on each single trip data, low-quality data is eliminated based on factors such as data integrity, accuracy and environmental variables, ensuring that the subsequently processed data is reliable and useful, avoiding interference in the subsequent map stitching process, and improving the quality and reliability of the map.

[0084] In some embodiments, Figure 4 As shown, the above step S230 further includes steps S231-S232, wherein at least part of the scan data in the multiple groups of scan data are screened based on the maximum clique algorithm according to the correlation of the multiple predicted scan trajectories in the geographical and / or temporal dimensions.

[0085] In step S231 , an undirected graph is constructed based on the correlation of a plurality of predicted scanning trajectories corresponding to at least part of the scanning data in geographical and / or temporal dimensions.

[0086] Each node in the undirected graph represents a predicted scanning trajectory, and an edge in the undirected graph represents that the correlation of the predicted scanning trajectories corresponding to two nodes in the geographical and / or time dimensions is greater than a second correlation threshold.

[0087] The correlation between two predicted scanning trajectories in the geographic dimension can be understood as the degree of similarity between the two predicted scanning trajectories. For example, the closer the starting positions of the two predicted scanning trajectories are, the higher the correlation is; the closer the midpoints of the two predicted scanning trajectories are, the higher the correlation is.

[0088] In step S232, based on the maximum clique algorithm, multiple maximum cliques are determined in the undirected graph, each of which includes multiple groups of target scanning data.

[0089] The maximum clique algorithm is a fast search algorithm that is used to find the maximum clique with the largest number of vertices in an undirected graph consisting of multiple points. Based on the correlation of the predicted trajectories of the scanned data, the maximum clique algorithm in graph theory is used to identify and aggregate single-trip data with high correlation. In this process, each identified maximum clique contains trajectories that are closely related to each other. The maximum clique found represents the set of trajectories that should be aggregated the most. Such aggregation can significantly reduce the need to process redundant data, thereby improving the overall data processing efficiency and system performance.

[0090] In some embodiments, the aforementioned step S240 of tracking the target features in each set of target data and determining the position and posture information corresponding to the multiple sets of target data in each scan data set further includes:

[0091] A plurality of target points corresponding to the target features in each set of target data are tracked, and the translation and rotation relationship of each set of target scanning data relative to the target features is determined.

[0092] According to the above technical means, by identifying the target features of the intersection environment, the extracted features are used to construct a detailed local map for each trip to provide sufficient details to support accurate matching.

[0093] In some embodiments, the aforementioned step S240, tracking the target features in each set of target data, and determining the posture information corresponding to multiple sets of target data in each scanning data set further includes: tracking multiple feature points corresponding to the target features in each set of the target scanning data, and determining the position deviation and angle deviation between different target data according to the different shapes, sizes and angles of the target features in different target scanning data.

[0094] According to the previous description, the target features are features with high reflectivity such as road signs in the intersection environment, which appear as a dense area in the point cloud. By tracking this part of the point cloud, the position and angle deviations between different point cloud data are determined based on the different shapes, sizes and angles of the same target features in different point cloud data, thereby aligning the data from different trips in space.

[0095] According to the above technical means, high-reflectivity target features are used as a more stable feature source. By identifying the same feature points in data from different passes, it is ensured that multiple passes of data can be correctly aligned in space, thus overcoming the instability problem of traditional methods based on surface and line features in multiple scans.

[0096] In some embodiments, Figure 5 As shown, the aforementioned step S250, according to the posture information corresponding to each group of target scanning data, splices multiple groups of target scanning data in multiple maximum clusters, and further includes steps S251-S253.

[0097] In step S251, a pose graph is constructed according to the pose information corresponding to each group of the target scan data.

[0098] It should be noted that the pose graph mentioned here is a pose graph constructed based on the spatial relationship within and between passes, including the poses of all single passes and the relative poses between passes.

[0099] In step S252, the pose of each target scan data is adjusted based on a pose graph optimization algorithm.

[0100] In step S253, the three-dimensional map of the intersection environment is spliced ​​according to the adjusted posture.

[0101] According to the above technical means, the posture of each single trip is accurately adjusted based on graph optimization technology, and then spliced ​​into a continuous and seamless point cloud map according to the optimized posture, which can ensure the overall consistency and accuracy of the map.

[0102] Combination of the above Figure 1-Figure 5 The method embodiment of the present application is described in detail. Figure 6 and Figure 7 , describing the device embodiment of the present application. It should be understood that the description of the device embodiment corresponds to the method embodiment, and therefore, the parts not described in detail can refer to the method embodiment above.

[0103] Figure 6 is a schematic structural diagram of a point cloud data processing device provided in an embodiment of the present application, Figure 6 The point cloud data processing device 600 in the embodiment includes:

[0104] The acquisition unit 610 is used to acquire multiple sets of scanning data, where the multiple sets of scanning data are laser point cloud data of the intersection environment collected by the vehicle in multiple scanning passes.

[0105] The first determination unit 620 is used to determine the predicted scanning trajectory of the scanning pass corresponding to each group of scanning data and the position information of each scanning pass.

[0106] The screening unit 630 is used to screen at least part of the multiple groups of scan data based on the maximum cluster algorithm according to the correlation of multiple predicted scan trajectories in geographical and / or time dimensions to obtain multiple maximum clusters; wherein each maximum cluster includes multiple groups of target scan data.

[0107] The second determination unit 640 is used to track the target features in each set of target scanning data and determine the position information corresponding to the multiple sets of target scanning data in each scanning data set, wherein the target features are high reflectivity markers in the intersection environment.

[0108] The stitching unit 650 is used to stitch multiple groups of target scanning data in multiple maximum clusters according to the posture information corresponding to each group of target scanning data to obtain a three-dimensional map of the intersection environment.

[0109] In some embodiments, the first determination unit 620 is further used to: process each set of scanning data using a laser inertial odometer (LIO) algorithm to generate a predicted scanning trajectory for each scanning pass; and use the coordinate information of the starting position of each scanning pass determined by a satellite positioning system as position information.

[0110] In some embodiments, the first determination unit 620 is further used to: determine a first predicted driving trajectory of the vehicle in each scanning pass based on an inertial measurement unit of the vehicle, and determine a second predicted driving trajectory of the vehicle in each scanning pass based on a satellite positioning system.

[0111] The screening unit 630 is further configured to screen the plurality of scan data sets according to the trajectory correlations of the first predicted driving trajectory, the second predicted driving trajectory, and the predicted scanning trajectory, and remove scan data sets whose trajectory correlations are less than a first correlation threshold.

[0112] In some embodiments, the screening unit 630 is also used to: construct an undirected graph based on the correlation of multiple predicted scanning trajectories corresponding to at least part of the scanning data in the geographical and / or time dimensions; wherein each node in the undirected graph represents a predicted scanning trajectory, and the edge in the undirected graph represents that the correlation of the predicted scanning trajectories corresponding to two nodes in the geographical and / or time dimensions is greater than a second correlation threshold; based on the maximum clique algorithm, determine multiple maximum clusters in the undirected graph, wherein each maximum cluster includes multiple groups of target scanning data.

[0113] In some embodiments, the second determination unit 640 is further used to: track multiple feature points corresponding to the target feature in each set of target scanning data, and determine the translation and rotation relationship of each set of target scanning data relative to the target feature.

[0114] In some embodiments, the target feature is a road sign in an intersection environment.

[0115] In some embodiments, the stitching unit 650 is also used to: construct a pose graph based on the pose information corresponding to each group of target scanning data; adjust the pose of each target scanning data based on the pose graph optimization algorithm; and stitch the three-dimensional map of the intersection environment based on the adjusted pose.

[0116] Figure 7 It is a schematic structural diagram of the point cloud data processing device provided in an embodiment of the present application. Figure 7 The dotted line in the figure indicates that the unit or module is optional. The point cloud data processing device 700 can be used to implement the method described in the above method embodiment. The point cloud data processing device 700 can be, for example, a server; or, the point cloud data processing device 700 can also be a chip or integrated circuit capable of implementing the method of the above embodiment.

[0117] The point cloud data processing device 700 may include one or more processors 710. The processor 710 may support the point cloud data processing device 700 to implement the method described in the foregoing method embodiment. The processor 710 may be a general-purpose processor or a special-purpose processor. For example, the processor may be a central processing unit (CPU). Alternatively, the processor 710 may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0118] The point cloud data processing device 700 may further include one or more memories 720. The memory 720 stores a computer program, which can be executed by the processor 710, so that the processor 710 executes the method described in the above method embodiment. The memory 720 may be independent of the processor 710 or integrated in the processor 710.

[0119] The point cloud data processing device 700 may further include a transceiver 730. The processor 710 may communicate with other devices or chips through the transceiver 730. For example, the processor 710 may transmit and receive data with other devices or chips through the transceiver 730.

[0120] The present application also provides a chip in an embodiment, including a processor, which can be used to call and run a computer program from a memory, so that a device equipped with the chip performs the method described in the above method embodiment. It is understood that the processor can be any type of processor mentioned above. It is understood that the memory can be independent of the chip or integrated in the chip.

[0121] The embodiments of the present application also provide a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the methods in the embodiments of the present application can be implemented.

[0122] The embodiment of the present application also provides a computer program product. The computer program product includes a program, which enables a computer to execute the method in each embodiment of the present application.

[0123] The present application also provides a computer program that enables a computer to execute the methods in the various embodiments of the present application.

[0124] It should be understood that in the embodiment of the present application, "B corresponding to A" means that B is associated with A, and B can be determined according to A. However, it should also be understood that determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information.

[0125] It should be understood that the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0126] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0127] In the several embodiments provided in the present application, it should be understood that the disclosed systems and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0128] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0129] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0130] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loading and executing computer program instructions on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber Line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can read or a data storage device such as a server, a data center, etc. that contains one or more available media integrated. Available media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, digital video discs (DVD)), or semiconductor media (eg, solid state disks (SSD)).

[0131] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A point cloud data processing method, characterized in that: The method comprises: Acquire multiple sets of scanning data, where the multiple sets of scanning data are laser point cloud data of the intersection environment collected by the vehicle in multiple scanning passes; Determine a predicted scanning trajectory of a scanning pass corresponding to each group of the scanning data and position information of each scanning pass; According to the correlation of the plurality of predicted scanning trajectories in the geographical and / or temporal dimensions, at least part of the scanning data in the plurality of groups of scanning data is screened based on a maximum clique algorithm to obtain a plurality of maximum cliques; wherein each of the maximum cliques includes a plurality of groups of target scanning data; Tracking the target features in each set of the target scanning data, and determining the position information corresponding to the multiple sets of target scanning data in each of the scanning data sets; wherein the target features are high-reflectivity markers in the intersection environment; According to the posture information corresponding to each group of the target scan data, multiple groups of the target scan data in the multiple largest groups are spliced ​​to obtain a three-dimensional map of the intersection environment.

2. The method according to claim 1, characterized in that The step of determining a predicted scanning trajectory of a scanning pass corresponding to each group of the scanning data and position information of each scanning pass includes: Each group of the scanning data is processed by using a laser inertial odometry method to generate a predicted scanning trajectory for each scanning pass; and the coordinate information of the starting position of each scanning pass is determined by a satellite positioning system as the position information.

3. The method according to claim 2, characterized in that Before screening the plurality of scan data sets based on the maximum clique algorithm according to the correlation of the plurality of predicted trajectories in geographical and / or temporal dimensions, the method further comprises: Determining a first predicted driving trajectory of the vehicle in each of the scanning passes based on an inertial measurement unit of the vehicle; Determining a second predicted driving trajectory of the vehicle in each of the scanning passes based on the satellite positioning system; The plurality of groups of scanning data are screened according to the trajectory correlations among the first predicted driving trajectory, the second predicted driving trajectory, and the predicted scanning trajectory, and scanning data whose trajectory correlations are less than a first correlation threshold are eliminated.

4. The method according to any one of claims 1 to 3, characterized in that The screening of at least part of the scan data in the plurality of groups of scan data based on the maximum clique algorithm comprises: constructing an undirected graph based on the correlation of the plurality of predicted scan trajectories corresponding to the at least part of the scan data in geographical and / or temporal dimensions; Each node in the undirected graph represents one of the predicted scanning trajectories, and an edge in the undirected graph represents that the correlation of the predicted scanning trajectories corresponding to two nodes in the geographical and / or time dimensions is greater than a second correlation threshold; Based on the maximum clique algorithm, a plurality of maximum cliques are determined in the undirected graph, each of the maximum cliques including a plurality of groups of the target scanning data.

5. The method according to claim 1, characterized in that Tracking the target features in each set of the target scan data and determining the pose information corresponding to the multiple sets of target scan data in each of the scan data sets includes: A plurality of feature points corresponding to the target feature in each set of the target scanning data are tracked to determine the translation and rotation relationship of each set of the target scanning data relative to the target feature.

6. The method according to claim 5, characterized in that Tracking the target features in each set of the target scan data and determining the position and posture information corresponding to the multiple sets of target scan data in each of the scan data sets also includes: Tracking a plurality of feature points corresponding to the target features in each set of the target scanning data, and determining the position deviation and angle deviation between different target data according to the differences in shapes, sizes and angles of the target features in different target scanning data; The target feature is a road sign in the intersection environment.

7. The method according to any one of claims 1 to 3, 5 or 6, characterized in that The step of splicing the multiple groups of target scan data in the multiple largest groups according to the pose information corresponding to each group of the target scan data comprises: Constructing a pose graph according to the pose information corresponding to each group of the target scanning data; Adjusting the pose of each target scan data based on a pose graph optimization algorithm; According to the adjusted position and posture, a three-dimensional map of the intersection environment is stitched.

8. A point cloud data processing device, characterized in that: The device comprises: An acquisition unit, used to acquire multiple sets of scanning data, wherein the multiple sets of scanning data are laser point cloud data of the intersection environment collected by the vehicle in multiple scanning passes; A first determining unit, used to determine a predicted scanning trajectory of a scanning pass corresponding to each group of the scanning data and position information of each scanning pass; A screening unit, configured to screen at least part of the plurality of groups of scan data based on a maximum cluster algorithm according to the correlation of the plurality of predicted scan trajectories in geographical and / or temporal dimensions, to obtain a plurality of maximum clusters; wherein each of the maximum clusters includes a plurality of groups of target scan data; A second determination unit is used to track the target feature in each group of the target scanning data, and determine the position information corresponding to the multiple groups of target scanning data in each of the scanning data sets, wherein the target feature is a high-reflectivity marker in the intersection environment; A stitching unit is used to stitch multiple groups of target scan data in the multiple largest groups according to the posture information corresponding to each group of target scan data, so as to obtain a three-dimensional map of the intersection environment.

9. A point cloud data processing device, characterized in that: The point cloud data processing device comprises a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium is used to store a computer program, and when the computer program is executed, the method according to any one of claims 1 to 7 is implemented.

Images