A Method for Optimizing the Precision of Single-line LiDAR Point Cloud Constrained by Multi-line LiDAR

Abstract

The present invention provides a method, system, electronic device and storage medium for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar. The method includes the following steps: S1, converting the single-line point cloud data and the multi-line point cloud data into the WGS84 coordinate system, separating the single-line point cloud data into equally spaced point cloud blocks, and fusing the point cloud blocks with the multi-line point cloud data; S2, inputting the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information; S3, taking the point cloud blocks as units, vertically correcting the single-line point cloud data through the road surface information in the multi-line point cloud data; horizontally correcting the single-line point cloud data through the sign information in the multi-line point cloud data; and traversing all point cloud blocks. The present invention utilizes the high-quality relative accuracy of the multi-line lidar to optimize the jump problem of the single-frame point cloud data of the single-line lidar, which helps to improve the quality of the automatic extraction results of high-precision map elements based on the point cloud data.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving high-precision map making, and more specifically, to a method, system, electronic device and storage medium for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar. Background Art

[0002] As an indispensable and important part of autonomous driving vehicles, the autonomous driving high-precision map provides favorable support for vehicle positioning, path planning, vehicle energy conservation, etc. To ensure the freshness of the high-precision map, how to achieve low-cost and high-efficiency map making and updating while meeting the accuracy requirements has become the key to ensuring the effectiveness and competitiveness of the map. The autonomous driving high-precision map update has the following characteristics: 1) The autonomous driving high-precision map is different from the traditional navigation map. It contains three-dimensional information and has higher accuracy requirements; 2) The high-precision map is a lane-level map; 3) The map information is richer and the update difficulty is higher.

[0003] In summary, the production and update process of the autonomous driving high-precision map needs to ensure high precision and high efficiency. In terms of improving efficiency, we can start from the perspective of reducing human intervention and use deep learning to improve efficiency. Deep learning technology in image processing is relatively mature, but images can only express two-dimensional information. Therefore, the point cloud data is dimensionally reduced, processed into a grayscale image for model inference, and then the result is back-calculated into the three-dimensional space to achieve the effect of complementary advantages. However, before the dimensionality reduction processing of the point cloud data, data jump problems are likely to occur, resulting in distorted and deformed point cloud data. In view of this, a solution for optimizing the quality of point cloud data needs to be designed. Summary of the Invention

[0004] The present invention aims at the technical problems existing in the prior art, and provides a method, system, electronic device and storage medium for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar, which uses the relatively high-quality relative accuracy of the multi-line lidar to optimize the data jump problem of the single-frame point cloud data of the single-line lidar, ensures that the data quality meets the requirements, and helps to improve the quality of the automatic extraction results of high-precision map elements based on point cloud data.

[0005] According to the first aspect of the present invention, a method for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar is provided, including the following steps:

[0006] S1, converting the single-line point cloud data and the multi-line point cloud data into the WGS84 coordinate system, separating the single-line point cloud data into equally spaced point cloud blocks, and fusing the point cloud blocks with the multi-line point cloud data;

[0007] S2, inputting the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information;

[0008] S3. Taking the point cloud block as a unit, vertically correct the single-line point cloud data through the road surface information in the multi-line point cloud data; horizontally correct the single-line point cloud data through the sign information in the multi-line point cloud data; traverse all point cloud blocks.

[0009] Based on the above technical solutions, the present invention can also be improved as follows.

[0010] Optionally, step S1 includes:

[0011] S101. Collect the original data, where the original data includes single-line point cloud data and multi-line point cloud data, and convert all points in each frame of the original data to a relative coordinate system with the acquisition point as the origin;

[0012] S102. Add the GPS coordinate parameters to all points in each frame of data in the relative coordinate system and convert it to the WGS84 coordinate system;

[0013] S103. Preset a length threshold s, equally spaced cut the single-line point cloud data in the WGS84 coordinate system according to the length threshold s to generate a number of point cloud blocks, and fuse each point cloud block with the multi-line point cloud data.

[0014] Optionally, in step S102, it further includes:

[0015] For sections where the GPS signal is lost, use the IMU inertial data and the stable GPS trajectory data at both ends of the section to interpolate the GPS trajectory of this section to obtain complete point cloud data based on the GPS coordinate system.

[0016] Optionally, the interpolation of the GPS trajectory of this section using the IMU inertial data and the stable GPS trajectory data at both ends of the section includes:

[0017] Starting from the starting end where the GPS signal is lost, obtain the distance traveled by the vehicle according to the acceleration and angular velocity of adjacent frames in the IMU inertial data, and then determine the direction of the vehicle's real-time movement according to the orientation in the IMU inertial data, so as to obtain the estimated driving trajectory of the vehicle on this section;

[0018] Optimize the estimated driving trajectory of this section by using curve fitting interpolation with the GPS position information at both ends of the section where the GPS signal is lost.

[0019] Optionally, in step S103, the fusion of each point cloud block with the multi-line point cloud data includes:

[0020] Centering on the middle frame of the point cloud block, select the frame with the shortest distance relative to the center from the multi-line point cloud data, and match the middle frame of the point cloud block with the frame with the shortest distance in the multi-line point cloud data; traverse all point cloud blocks until the fusion of all point cloud blocks of the single-line point cloud data and the multi-line point cloud data is achieved.

[0021] Optionally, in step S103, the basis for setting the length threshold s is the scanning distance parameter of the multi-line lidar device and the single-frame point density distribution.

[0022] Optionally, step S3 includes:

[0023] S301, find the road surface part from the road surface information extracted from a single point cloud block, calculate the length L of the road surface part in the single-line point cloud data, and calculate the average elevation H of all points corresponding to the road surface part in the single-line point cloud data;

[0024] Calculate the distances L1 and L2 between the front and rear frames and the current frame in the single-line point cloud data respectively. Use L1 + L2 as the width and L as the length to form a rectangular frame R. Obtain the average elevation H1 of all points within the multi-line point cloud data corresponding to the rectangular frame R. Calculate the elevation difference |H1 - H| through the following formula, and compare it with the empirical value e of the elevation difference, where e is a non-zero constant, and the value of e is related to the performance of the lidar of the sampled point cloud data; when |H1 - H| > e and e > 0, correct the elevation of each point in the single-line point cloud data of the current frame vertically according to the elevation difference.

[0025] S302, if the sign information is extracted from the current point cloud block in step S2, the corresponding sign information in the multi-line point cloud data includes the left and right edges of the sign, and the corresponding sign information in the single-line point cloud data includes the line segment formed by the sign part; align the line segment formed by the sign part in the corresponding frame of the single-line point cloud data with the left and right edges of the sign in the multi-line point cloud data, and calculate the horizontal displacement difference w before and after the alignment of the line segment, and compare it with the empirical value e1 of the horizontal displacement difference, where e1 is a non-zero constant, and the value of e1 is related to the performance of the lidar of the sampled point cloud data; when w > e1 and e1 > 0, correct each point in several frames of the single-line point cloud data where the sign is located horizontally according to the value of w.

[0026] S303, traverse all point cloud blocks until the correction of the single-line point cloud data corresponding to all point cloud blocks is completed.

[0027] According to the second aspect of the present invention, there is provided a single-line laser point cloud accuracy optimization system constrained by a multi-line lidar, including:

[0028] A data fusion module, configured to convert single-line point cloud data and multi-line point cloud data into the WGS84 coordinate system, separate the single-line point cloud data into equidistant point cloud blocks, and fuse the point cloud blocks with the multi-line point cloud data;

[0029] An information extraction module, configured to input the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information;

[0030] A data correction module, configured to perform vertical correction on the single-line point cloud data by using the road surface information in the multi-line point cloud data in units of point cloud blocks; perform horizontal correction on the single-line point cloud data by using the sign information in the multi-line point cloud data; and traverse all point cloud blocks.

[0031] According to a third aspect of the present invention, there is provided an electronic device, including a memory and a processor, and the processor is configured to implement the steps of a method for optimizing the accuracy of single-line laser point cloud constrained by a multi-line lidar when executing a computer management program stored in the memory.

[0032] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium, on which a computer management program is stored, and the computer management program is configured to implement the steps of a method for optimizing the accuracy of single-line laser point cloud constrained by a multi-line lidar when executed by a processor.

[0033] A method, system, electronic device, and storage medium for optimizing the accuracy of single-line laser point cloud constrained by a multi-line lidar provided by the present invention utilize the relatively high accuracy of the multi-line lidar point cloud to constrain / correct the quality of the local adjacent frame point cloud data of the single-line lidar, optimize the jump problem of the single-frame point cloud data of the single-line lidar, so as to achieve the purpose of improving the quality of the single-line lidar point cloud data. The present invention improves the quality of the single-line lidar point cloud data to improve the input data quality for automatic extraction of traffic elements based on point cloud data and the construction of point cloud maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a principle block diagram and flowchart of a method for optimizing the accuracy of single-line laser point cloud constrained by a multi-line lidar provided by the present invention;

[0035] Figure 2 It is a flowchart of a method for optimizing the accuracy of single-line laser point cloud constrained by a multi-line lidar provided by the present invention;

[0036] Figure 3 It is a flowchart of step S1 of the present invention;

[0037] Figure 4 It is a flowchart of step S3 of the present invention;

[0038] Figure 5The simulation diagram of part of the point cloud data after fusion of the present invention;

[0039] Figure 6 is Figure 5 The schematic diagram of the local magnification principle in

[0040] Figure 7 The composition structure diagram of a single-line laser point cloud accuracy optimization system constrained by a multi-line lidar provided by the present invention;

[0041] Figure 8 The schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0042] Figure 9 The schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. Detailed implementation manners

[0043] The following combines the drawings and embodiments to further describe in detail the specific implementation manners of the present invention. The following embodiments are used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0044] Figure 1 and Figure 2 respectively are the principle diagram and flowchart of a single-line laser point cloud accuracy optimization method constrained by a multi-line lidar provided by the present invention. As shown in Figure 1 and Figure 2 shown, the method includes:

[0045] S1. Convert the single-line point cloud data and the multi-line point cloud data to the WGS84 coordinate system, divide the single-line point cloud data into equally spaced point cloud blocks, and fuse the point cloud blocks with the multi-line point cloud data;

[0046] S2. Input the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information;

[0047] S3. Taking the point cloud blocks as units, vertically correct the single-line point cloud data through the road surface information in the multi-line point cloud data; horizontally correct the single-line point cloud data through the sign information in the multi-line point cloud data; traverse all point cloud blocks.

[0048] It can be understood that based on the defects in the background technology, the embodiment of the present invention proposes a single-line laser point cloud accuracy optimization method constrained by a multi-line lidar. This method uses the relatively high accuracy of the multi-line lidar point cloud to constrain / correct the quality of the local adjacent frame point cloud data of the single-line lidar, and optimizes the jump problem of the single-frame point cloud data of the single-line lidar, so as to achieve the purpose of improving the quality of the single-line lidar point cloud data. The present invention improves the quality of the single-line lidar point cloud data to improve the input data quality of the automatic extraction of traffic elements based on the point cloud data and the construction of the point cloud map.

[0049] In a possible embodiment, as Figure 3 shown, step S1 includes:

[0050] S101, collect raw data, where the raw data at least includes single-line point cloud data and multi-line point cloud data, and convert all points in each frame of raw data to a relative coordinate system with the acquisition point as the origin;

[0051] S102, add GPS coordinate parameters to all points in each frame of data in the relative coordinate system and convert it to the WGS84 coordinate system;

[0052] S103, preset a length threshold s, equally space cut the single-line point cloud data in the WGS84 coordinate system according to the length threshold s to generate several point cloud blocks, and fuse each point cloud block with the multi-line point cloud data.

[0053] It can be understood that in this embodiment, the multi-line point cloud data and the single-line point cloud data are respectively resolved, and through data resolution, the multi-line point cloud data and the single-line point cloud data are respectively converted to the WGS84 coordinate system and data fusion is performed. In this embodiment, the single-line laser point cloud data is also cut into several point cloud blocks, and the point cloud blocks are used as the basic processing units for subsequent road information / signboard information extraction and data correction.

[0054] In a possible embodiment, in step S102, it further includes:

[0055] For sections where the GPS signal is lost, use IMU inertial data and stable GPS trajectory data at both ends of the section to interpolate the GPS trajectory of this section to obtain complete point cloud data based on the GPS coordinate system.

[0056] It can be understood that the collected raw data at least further includes IMU inertial data and GPS trajectory data, and the data in the GPS coordinate system of the section where the GPS signal is lost is improved by the interpolation method to smooth the vehicle driving trajectory.

[0057] In a possible embodiment, the interpolation of the GPS trajectory of this section using IMU inertial data and stable GPS trajectory data at both ends of the section includes:

[0058] Starting from the starting end where the GPS signal is lost, obtain the distance the vehicle moves according to the acceleration and angular velocity of adjacent frames in the IMU inertial data, and then determine the direction of the vehicle's real-time movement according to the orientation in the IMU inertial data, so as to obtain the estimated driving trajectory of the vehicle on this section;

[0059] The estimated driving trajectory of the section is optimized by using curve fitting interpolation with the GPS position information at both ends of the section where the GPS signal is lost.

[0060] It can be understood that IMU inertial data is collected through an IMU inertial navigation system. The IMU inertial navigation system at least includes an accelerometer (to estimate the vehicle's driving acceleration), a gyroscope (to estimate the angular velocity of the X, Y, and Z axes), and a magnetometer (to estimate the vehicle's head orientation). During the vehicle's driving process, corresponding data will be collected at a certain sampling frequency. Starting from the starting end where the GPS signal is lost, the distance that the vehicle moves between adjacent frames of IMU inertial data is obtained based on the acceleration and angular velocity, and then the real-time moving direction of the vehicle is determined according to the head orientation, so as to estimate the driving trajectory of the vehicle on this section. However, there are errors in the estimated value. At the end of the section where the GPS signal is lost, there will be a deviation between the vehicle position estimated based on the IMU inertial data and the actual position of the vehicle. Therefore, it is necessary to optimize the trajectory estimated value of this section by using curve fitting interpolation with the GPS position information at both ends.

[0061] In a possible implementation manner, in step S103, the fusion of each point cloud block with the multi-line point cloud data includes:

[0062] Taking the middle frame of the point cloud block as the center, selecting the frame with the closest distance to the center in the multi-line point cloud data, and matching the middle frame of the point cloud block with the frame with the closest distance in the multi-line point cloud data; traversing all point cloud blocks until the fusion of all point cloud blocks of the single-line point cloud data and the multi-line point cloud data is achieved.

[0063] It can be understood that achieving data fusion is to prepare for subsequent correction of the single-line point cloud data using the multi-line point cloud data.

[0064] In a possible implementation manner, in step S103, the basis for setting the length threshold s is the scanning distance parameter of the multi-line lidar device and the single-frame point density distribution.

[0065] It can be understood that setting an appropriate length threshold s according to the actual situation of the point cloud data acquisition device to cut the single-line point cloud data can ensure the efficiency of subsequent data correction steps.

[0066] In a possible implementation manner, as Figure 4 shown, step S3 includes the following steps:

[0067] S301, find the road surface part from the road surface information extracted from a single point cloud block, calculate the length L of the road surface part in the single-line point cloud data, and calculate the average elevation H of all points corresponding to the road surface part in the single-line point cloud data.

[0068] As Figure 5The figure shows a simulation diagram of some of the fused point cloud data. Figure 5 The solid rectangular frame in the figure shows the road surface part, and the dashed parts marked 1, 2, and 3 are any three single-line point cloud data. As Figure 5 shown, the road width is used as the length L in this embodiment. Each single-line point cloud data is composed of several points, and each point has its own elevation information. The average elevation H in this embodiment corresponds to the average elevation of all points in any one of the single-line point cloud data such as 1, 2, and 3.

[0069] Figure 6 For Figure 5 the simplified local schematic diagram, calculate the distances L1 and L2 between the front and rear frames and the current frame in the single-line point cloud data. Assume Figure 6 that the single-line point cloud data marked 2 in the figure is the current frame, the single-line point cloud data marked 1 is the front frame, and the single-line point cloud data marked 3 is the rear frame. Then L1 is the distance between the single-line point cloud data marked 1 and the single-line point cloud data marked 2. Similarly, L2 is the distance between the single-line point cloud data marked 3 and the single-line point cloud data marked 2. A rectangular frame R is formed with L1 + L2 as the width and L as the length. The rectangular frame R is as shown by the dashed frame in Figure 6 Then obtain the average elevation H1 of all points in the multi-line point cloud data corresponding to the rectangular frame R, and calculate the elevation difference H1 - H through the following formula, and compare it with the empirical value e of the elevation difference, where e is a non-zero constant, and the value of e is related to the performance of the lidar for sampling the point cloud data; when |H1 - H| > e and e > 0, correct the elevation of each point of the single-line point cloud data of the current frame in the vertical direction according to the elevation difference;

[0070] S302. If the sign information is extracted from the current point cloud block in step S2, the corresponding sign information in the multi-line point cloud data includes the left and right edges of the sign, and the corresponding sign information in the single-line point cloud data includes the line segment formed by the sign part; align the line segment formed by the sign part of the corresponding frame in the single-line point cloud data according to the left and right edges of the sign in the multi-line point cloud data, and calculate the horizontal displacement difference w before and after the alignment of the line segment, and compare it with the empirical value e1 of the horizontal displacement difference, where e1 is a non-zero constant, and the value of e1 is related to the performance of the lidar for sampling the point cloud data; when w > e1 and e1 > 0, correct each point of the single-line point cloud data of several frames where this sign is located in the horizontal direction according to the value of w;

[0071] S303. Traverse all point cloud blocks until the correction of the single-line point cloud data corresponding to all point cloud blocks is completed.

[0072] It can be understood that in step S3, based on the road surface information and sign information extracted in step S2, multi-line point cloud data is used to correct the single-line point cloud data in the vertical direction and the horizontal direction to improve the accuracy of the single-line point cloud data. Step S3 takes a single point cloud block as the processing unit. After processing one point cloud block, the next point cloud block is processed until all point cloud blocks are traversed. For the point cloud block where only road surface information is recognized and sign information is not recognized, only elevation correction, that is, vertical direction correction, is required; for the point cloud block with both road surface information and sign information, vertical direction correction is performed first, and then horizontal direction correction is performed.

[0073] Figure 7 The figure is a structural diagram of a single-line laser point cloud accuracy optimization system constrained by a multi-line lidar provided by an embodiment of the present invention. As Figure 7 shown, a single-line laser point cloud accuracy optimization system constrained by a multi-line lidar includes a data fusion module 101, an information extraction module 102, and a data correction module 103, where:

[0074] The data fusion module 101 is used to convert the single-line point cloud data and the multi-line point cloud data into the WGS84 coordinate system, divide the single-line point cloud data into equally spaced point cloud blocks, and fuse the point cloud blocks with the multi-line point cloud data;

[0075] The information extraction module 102 is used to input the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information;

[0076] The data correction module 103 is used to take the point cloud block as a unit, perform vertical correction on the single-line point cloud data through the road surface information in the multi-line point cloud data; perform horizontal correction on the single-line point cloud data through the sign information in the multi-line point cloud data; traverse all point cloud blocks.

[0077] It can be understood that a single-line laser point cloud accuracy optimization system constrained by a multi-line lidar provided by the present invention corresponds to the multi-line lidar constrained single-line laser point cloud accuracy optimization method provided by the foregoing embodiments. The relevant technical features of the single-line laser point cloud accuracy optimization system constrained by a multi-line lidar can refer to the relevant technical features of the multi-line lidar constrained single-line laser point cloud accuracy optimization method, which will not be elaborated here.

[0078] Please refer to Figure 8 , Figure 8 which is a schematic diagram of an embodiment of an electronic device provided by an embodiment of the present invention. As Figure 8 shown, an embodiment of the present invention provides an electronic device 800, including a memory 810, a processor 820, and a computer program 811 stored in the memory 810 and executable on the processor 820. When the processor 820 executes the computer program 811, the following steps are implemented:

[0079] Convert the single-line point cloud data and multi-line point cloud data to the WGS84 coordinate system, divide the single-line point cloud data into equally spaced point cloud blocks, and fuse the point cloud blocks with the multi-line point cloud data;

[0080] Input the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information;

[0081] Taking the point cloud blocks as units, vertically correct the single-line point cloud data through the road surface information in the multi-line point cloud data; horizontally correct the single-line point cloud data through the sign information in the multi-line point cloud data; traverse all point cloud blocks.

[0082] Please refer to Figure 9 , Figure 9 which is a schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. As Figure 9 shown, this embodiment provides a computer-readable storage medium 900, on which a computer program 911 is stored. When the computer program 911 is executed by a processor, the following steps are implemented:

[0083] Convert the single-line point cloud data and multi-line point cloud data to the WGS84 coordinate system, divide the single-line point cloud data into equally spaced point cloud blocks, and fuse the point cloud blocks with the multi-line point cloud data;

[0084] Input the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information;

[0085] Taking the point cloud blocks as units, vertically correct the single-line point cloud data through the road surface information in the multi-line point cloud data; horizontally correct the single-line point cloud data through the sign information in the multi-line point cloud data; traverse all point cloud blocks.

[0086] A method, system and storage medium for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar provided by the embodiments of the present invention use the relatively high accuracy of the multi-line lidar point cloud to constrain / correct the quality of the local adjacent frame point cloud data of the single-line lidar, and optimize the problem of the jump of the single-frame point cloud data of the single-line lidar, so as to achieve the purpose of improving the quality of the single-line lidar point cloud data. The present invention improves the quality of the single-line lidar point cloud data to improve the input data quality of the automatic extraction of traffic elements based on the point cloud data and the construction of the point cloud map.

[0087] It should be noted that in the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0088] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0089] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the processes Figure 1 or blocks.

[0090] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the processes Figure 1 or blocks.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the processes Figure 1 or blocks.

[0092] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0093] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A method for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar, characterized in that It includes the following steps: S1. Convert the single-line point cloud data and multi-line point cloud data to the WGS84 coordinate system, divide the single-line point cloud data into equally spaced point cloud blocks, and fuse the point cloud blocks with the multi-line point cloud data; S2. Input the fused point cloud data into the trained semantic segmentation model to extract road surface information and sign information; S3. Taking the point cloud blocks as units, vertically correct the single-line point cloud data through the road surface information in the multi-line point cloud data; horizontally correct the single-line point cloud data through the sign information in the multi-line point cloud data; traverse all point cloud blocks; specifically including: S301. Find the road surface part from the road surface information extracted from a single point cloud block, calculate the length L of the road surface part in the single-line point cloud data, and calculate the average elevation H of all points corresponding to the road surface part in the single-line point cloud data; Calculate the distances L1 and L2 between the front and rear frames and the current frame in the single-line point cloud data respectively. Construct a rectangular frame R with L1 + L2 as the width and L as the length, and obtain the average elevation H1 of all points in the multi-line point cloud data corresponding to the rectangular frame R. Calculate the elevation difference through the following formula , and compare it with the empirical value e of the elevation difference, where e is a non-zero constant, and the value of e is related to the performance of the lidar of the sampled point cloud data; when satisfying , and e > 0, correct the elevation of each point in the single-line point cloud data of the current frame in the vertical direction according to the elevation difference; S302. If the sign information is extracted from the current point cloud block in step S2, the corresponding sign information in the multi-line point cloud data includes the left and right side lines of the sign, and the corresponding sign information in the single-line point cloud data includes the line segments formed by the sign part; align the line segments formed by the sign part in the corresponding frame of the single-line point cloud data with the left and right side lines of the sign in the multi-line point cloud data, calculate the horizontal displacement difference w before and after the alignment of the line segments, and compare it with the empirical value e1 of the horizontal displacement difference, where e1 is a non-zero constant, and the value of e1 is related to the performance of the lidar for sampling the point cloud data; when is satisfied and e1>0, correct each point of the single-line point cloud data of several frames where this sign is located in the horizontal direction according to the value of w. S303. Traverse all point cloud blocks until the single-line point cloud data corresponding to all point cloud blocks is corrected.

2. A method for optimizing the accuracy of a single-line laser point cloud constrained by a multi-line lidar according to claim 1, characterized in that Step S1 includes: S101. Collect the original data, where the original data includes single-line point cloud data and multi-line point cloud data, and convert all points in each frame of the original data to the relative coordinate system with the collection point as the origin; S102. Add GPS coordinate parameters to all points in each frame of data in the relative coordinate system and convert it to the WGS84 coordinate system; S103. Preset a length threshold s, cut the single-line point cloud data in the WGS84 coordinate system at equal intervals according to the length threshold s to generate several point cloud blocks, and fuse each point cloud block with the multi-line point cloud data.

3. A method for optimizing the accuracy of a single-line laser point cloud constrained by a multi-line lidar according to claim 2, characterized in that, In step S102, it also includes: For sections where the GPS signal is lost, use the IMU inertial data and the stable GPS trajectory data at both ends of the section to interpolate the GPS trajectory of this section to obtain complete point cloud data based on the GPS coordinate system.

4. A method for optimizing the accuracy of a single-line lidar point cloud constrained by a multi-line lidar according to claim 3, characterized in that, The interpolation of the GPS trajectory of this section using the IMU inertial data and the stable GPS trajectory data at both ends of the section includes: Starting from the starting end where the GPS signal is lost, obtain the distance traveled by the vehicle according to the acceleration and angular velocity of adjacent frames in the IMU inertial data, and then determine the real-time moving direction of the vehicle according to the orientation in the IMU inertial data, so as to obtain the estimated driving trajectory of the vehicle on this section; Optimize the estimated driving trajectory of this section by using curve fitting interpolation with the GPS position information at both ends of the section where the GPS signal is lost.

5. A method for optimizing the accuracy of a single-line lidar point cloud constrained by a multi-line lidar according to claim 2, characterized in that In step S103, the fusion of each point cloud block with the multi-line point cloud data includes: Taking the middle frame of the point cloud block as the center, select the frame with the closest distance relative to the center in the multi-line point cloud data, and match the middle frame of the point cloud block with the frame with the closest distance in the multi-line point cloud data; traverse all point cloud blocks until the fusion of all point cloud blocks of the single-line point cloud data and the multi-line point cloud data is achieved.

6. A method for optimizing the accuracy of a single-line laser point cloud constrained by a multi-line lidar according to claim 2, characterized in that, In step S103, the basis for setting the length threshold s is the scanning distance parameter of the multi-line lidar device and the single-frame point density distribution.

7. A single-line lidar point cloud accuracy optimization system constrained by multi-line lidar, characterized in that It includes: A data fusion module for converting the single-line point cloud data and multi-line point cloud data to the WGS84 coordinate system, dividing the single-line point cloud data into equally spaced point cloud blocks, and fusing the point cloud blocks with the multi-line point cloud data; An information extraction module, configured to input the fused point cloud data into a trained semantic segmentation model to extract road surface information and sign information; A data correction module, configured to take point cloud blocks as units, perform vertical correction on the single-line point cloud data through the road surface information in the multi-line point cloud data; perform horizontal correction on the single-line point cloud data through the sign information in the multi-line point cloud data; traverse all point cloud blocks; specifically including: Find the road surface part from the road surface information extracted from a single point cloud block, calculate the length L of the road surface part in the single-line point cloud data, and calculate the average elevation H of all points corresponding to the road surface part in the single-line point cloud data; Calculate the distances L1 and L2 between the front and rear frames and the current frame respectively in the single-line point cloud data. A rectangular frame R is formed with L1 + L2 as the width and L as the length. Obtain the average elevation H1 of all points in the multi-line point cloud data corresponding to the rectangular frame R, and calculate the elevation difference through the following formula , and compare it with the empirical value e of the elevation difference, where e is a non-zero constant, and the value of e is related to the performance of the lidar for sampling the point cloud data; when satisfying , and e > 0, correct the elevation of each point in the single-line point cloud data of the current frame in the vertical direction according to the elevation difference; If the signboard information is extracted from the current point cloud block in step S2, the corresponding signboard information in the multi-line point cloud data includes the left and right edges of the signboard, and the corresponding signboard information in the single-line point cloud data includes the line segments formed by the signboard part; align the line segments formed by the signboard part in the corresponding frame of the single-line point cloud data according to the left and right edges of the signboard in the multi-line point cloud data, calculate the horizontal displacement difference w before and after the alignment of the line segments, and compare it with the empirical value e1 of the horizontal displacement difference, where e1 is a non-zero constant, and the value of e1 is related to the performance of the lidar for sampling the point cloud data; when is satisfied and e1>0, correct each point of the single-line point cloud data of several frames where this signboard is located in the horizontal direction according to the value of w; Traverse all point cloud blocks until the correction of the single-line point cloud data corresponding to all point cloud blocks is completed.

8. An electronic device, characterized in that, It includes a memory and a processor, and the processor is configured to implement the steps of a method for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar as described in any one of claims 1-6 when executing a computer management program stored in the memory.

9. A computer-readable storage medium, characterized in that, A computer management program is stored thereon, and when the computer management program is executed by a processor, it implements the steps of a method for optimizing the accuracy of single-line laser point cloud constrained by multi-line lidar as described in any one of claims 1-6.

Images