Ground filtering method and device

Abstract

The embodiment of the present invention discloses a ground filtering method and device, which relates to the field of intelligent driving technology. The method includes: calibrating the acquisition device installed on the carrier, obtaining the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system; using the acquisition device to acquire the first single-frame point cloud data of the real-time environment of the carrier in the local coordinate system; using the transformation relationship to convert the coordinates of the first single-frame point cloud data from the local coordinate system to the absolute coordinate system; using an algorithm to establish a plane model; and filtering the first single-frame point cloud data according to the plane model and the elevation accuracy of the acquisition device. The ground filtering method of the embodiment of the present application improves the noise resistance, real-time and robustness of processing the first single-frame point cloud data.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to a ground filtering method and device. Background Art

[0002] In recent years, with the rapid development of intelligent driving technology, laser radar (LIDAR) has the characteristics of high accuracy, large ranging range, and no influence of light. It has been widely used in environmental perception fields such as obstacle detection, real-time positioning and map construction of intelligent driving vehicles.

[0003] Among the related technologies, one of the key technologies in intelligent driving technology is to use the on-board laser radar to collect point cloud data of the real-time environment, and to perceive the surrounding environment in real time based on the point cloud data to provide decisions for intelligent driving. The primary task to achieve real-time environmental perception is to perform ground filtering on the point cloud data to separate ground points from non-ground points. At present, the ground filtering technologies for on-board point cloud data mainly include: multi-scale progressive morphological filtering method and progressive encryption filtering method based on irregular triangulated network (TIN). However, the filtering results of the multi-scale progressive morphological filtering method are greatly affected by the setting of multi-parameter thresholds. A large number of experiments are required to obtain empirical thresholds for different laser radar equipment and installation methods. In addition, the method fails to filter abnormal noise points below the ground and has poor noise resistance. The progressive encryption filtering method of irregular triangulated network (TIN) requires iterative construction of triangulated network layer by layer. Affected by the slow speed of constructing triangulated network, multi-layer iterative processing makes the filtering method poor in timeliness and cannot meet the real-time requirements of intelligent driving. Summary of the invention

[0004] In order to overcome the problems in related technologies that ground filtering methods are greatly affected by multi-parameter threshold settings and have poor noise resistance and effectiveness, an embodiment of the present invention provides a ground filtering method and device, which improves the noise resistance, real-time and robustness of the ground filtering method in processing point cloud data.

[0005] According to one aspect of the present invention, there is provided a ground filtering method, comprising:

[0006] Calibrate the acquisition device installed on the carrier to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system;

[0007] Using the acquisition device to acquire first single-frame point cloud data of the real-time environment where the carrier is located in the local coordinate system;

[0008] Converting the coordinates of the first single-frame point cloud data from the local coordinate system to the absolute coordinate system using the transformation relationship;

[0009] Using algorithms to create plane models; and

[0010] The first single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the acquisition device.

[0011] Optionally, before establishing the plane model using an algorithm, the method includes:

[0012] Acquire a first height of the acquisition device installed on the carrier from the ground;

[0013] The coordinates of the acquisition device are converted from the local coordinate system to the absolute coordinate system using the transformation relationship.

[0014] Optionally, before establishing the plane model using an algorithm, the method further includes:

[0015] Calculate a second height of the ground in the absolute coordinate system according to the first height and the coordinates of the acquisition device in the absolute coordinate system;

[0016] The first single-frame point cloud data is filtered according to the second height and the first threshold to obtain second single-frame point cloud data.

[0017] Optionally, the establishing of the plane model by using an algorithm includes:

[0018] The second single-frame point cloud data is used to calculate mathematical model parameters of the second single-frame point cloud data using a random sampling consensus algorithm to obtain the plane model.

[0019] Optionally, filtering the first single-frame point cloud data according to the plane model and the elevation accuracy of the acquisition device includes:

[0020] Calculating the distance between the second single-frame point cloud data and the plane model;

[0021] When the distance is less than the negative value of the elevation accuracy, marking the second single-frame point cloud data as a noise point;

[0022] When the distance is greater than the elevation accuracy, marking the second single-frame point cloud data as a non-ground point; and

[0023] When the distance is greater than or equal to the negative value of the elevation accuracy and less than or equal to the elevation accuracy, the second single-frame point cloud data is marked as a ground point.

[0024] Optionally, calibrating the acquisition device installed on the carrier to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system includes:

[0025] Establishing a target normal equation group of the transformation relationship between the local coordinate system and the absolute coordinate system in the three-dimensional calibration field, and setting a plurality of control points in the three-dimensional calibration field;

[0026] Using a real-time differential positioning method to measure first coordinates of the plurality of control points in the absolute coordinate system;

[0027] In the three-dimensional calibration field, according to a preset route, using the acquisition device to acquire third single-frame point cloud data of the real-time environment where the carrier is located in the local coordinate system; and

[0028] Real-time dynamic posture information collected by integrated navigation equipment.

[0029] Optionally, the calibrating the acquisition device installed on the carrier to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system further includes:

[0030] Selecting the plurality of control points from the third single-frame point cloud data to obtain second coordinates of the plurality of control points in the local coordinate system;

[0031] Solving the target normal equations using a least squares algorithm according to the first coordinates, the second coordinates, and the position information of the plurality of control points to obtain the extrinsic parameters to be calibrated of the acquisition device; and

[0032] Based on the extrinsic parameters to be calibrated of the acquisition device and the target normal equation group, the transformation relationship between the local coordinate system and the absolute coordinate system is obtained.

[0033] Optionally, the target normal equation group of the transformation relationship between the local coordinate system and the absolute coordinate system is:

[0034]

[0035] in, is the second coordinate of the control point in the local coordinate system, is the first coordinate of the control point in the absolute coordinate system, is the real-time dynamic position information collected by the combined navigation device, is the rotation matrix of the integrated navigation device, is the coordinate of the integrated navigation device in the absolute coordinate system, It is the external parameter to be calibrated of the acquisition device.

[0036] According to another aspect of the present invention, there is provided a ground filtering device, comprising:

[0037] A calibration unit is configured to calibrate the acquisition device installed on the carrier to obtain a transformation relationship between a local coordinate system with the acquisition device as the center point coordinate and an absolute coordinate system;

[0038] A collection unit is configured to collect a first single-frame point cloud data of the real-time environment where the carrier is located in the local coordinate system by using the collection device;

[0039] A first coordinate transformation unit is configured to transform the coordinates of the first single-frame point cloud data from the local coordinate system to the absolute coordinate system by using the transformation relationship;

[0040] A building unit configured to build a plane model using an algorithm; and

[0041] The first filtering unit is configured to filter the first single-frame point cloud data according to the plane model and the elevation accuracy of the acquisition device.

[0042] Optionally, before the plane model is established by using an algorithm, the ground filtering device further includes:

[0043] An acquisition unit is configured to acquire a first height of the acquisition device mounted on the carrier from the ground;

[0044] The second coordinate transformation unit is configured to transform the coordinates of the acquisition device from the local coordinate system to the absolute coordinate system by using the transformation relationship.

[0045] Optionally, before the plane model is established by using an algorithm, the ground filtering device further includes:

[0046] A calculation unit, configured to calculate a second height of the ground in the absolute coordinate system according to the first height and the coordinates of the acquisition device in the absolute coordinate system;

[0047] The second filtering unit is configured to filter the first single-frame point cloud data according to the second height and the first threshold to obtain second single-frame point cloud data.

[0048] According to another aspect of the present invention, there is provided a ground filtering control device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the above-mentioned ground filtering method.

[0049] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed, the ground filtering method as described above is implemented.

[0050] According to another aspect of the present invention, a computer program product is provided, including a computer program product, wherein the computer program includes program instructions, and when the program instructions are executed by a mobile terminal, the mobile terminal executes the steps of the above-mentioned ground filtering method.

[0051] An embodiment of the present invention has the following advantages or beneficial effects:

[0052] The first single-frame point cloud data is used to calculate the mathematical model parameters of the first single-frame point cloud data using a random sampling consensus algorithm (RANSAC algorithm) to obtain a plane model. The first single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the laser radar. The first single-frame point cloud data is filtered and classified into ground points, non-ground points and noise points, thereby improving the anti-noise performance of the ground filtering method in processing the first single-frame point cloud data. By establishing a plane model, the first single-frame point cloud data collected by laser radars of different types and / or different installation methods is ground filtered, without the need to set multiple parameter thresholds, thereby improving the real-time and robustness of the ground filtering method in processing the first single-frame point cloud data. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0054] Figure 1 A schematic flow chart showing a ground filtering method according to an embodiment of the present invention.

[0055] Figure 2 A schematic flow chart showing a ground filtering method according to an embodiment of the present invention.

[0056] Figure 3 A schematic flow chart showing a calibration method for a collection device according to an embodiment of the present invention.

[0057] Figure 4 A schematic structural diagram of a ground filtering device according to an embodiment of the present invention is shown.

[0058] Figure 5 A schematic structural diagram of a ground filtering device according to an embodiment of the present invention is shown.

[0059] Figure 6 A schematic structural diagram of a ground filtering control device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0060] The present invention is described below based on embodiments, but the present invention is not limited to these embodiments. In the detailed description of the present invention below, some specific details are described in detail. It is possible for a person skilled in the art to fully understand the present invention without the description of these details. In order to avoid confusing the essence of the present invention, known methods, processes, and procedures are not described in detail. In addition, the drawings are not necessarily drawn to scale.

[0061] In recent years, with the invention and popularization of 3D laser sensors such as Kinect and Velodyne whose rotation axis is perpendicular to the ground, the research on 3D depth data processing methods has received more and more attention. The 3D point cloud data obtained by scanning and sampling the surrounding environment by Velodyne and other laser radar sensors contains the reflection data of almost all objects in the sensor's surrounding environment. The purpose of obstacle detection and identification can be achieved by processing the scanned 3D point cloud data. However, since the scale of 3D point cloud data is often too large, it will bring real-time problems to subsequent algorithms. Therefore, the first problem to be solved is to perform ground filtering and segmentation on the point cloud data to reduce the data scale. Among them, ground filtering refers to the identification and filtering of ground points and non-ground points in point cloud data.

[0062] The following uses the vehicle-mounted laser radar in intelligent driving technology as an application scenario to explain in detail how to use the ground filtering method to identify and filter out ground points and non-ground points in the point cloud data of the real-time environment.

[0063] Figure 1 FIG. 1 is a flow chart of a ground filtering method according to an embodiment of the present invention. Specifically, the following steps are included:

[0064] In step S110, the acquisition device installed on the carrier is calibrated to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system.

[0065] In this step, in the intelligent driving system, the laser radar can be installed on the top of the intelligent driving vehicle, and the position data information obtained by the laser radar is based on the radar's own local coordinate system. In the actual use of the laser radar, in order to unify the information of multiple sensors, it is necessary to convert the data obtained by the laser radar in its own local coordinate system to the absolute coordinate system. During the data conversion process, the point cloud data in the local coordinate system with the laser radar as the coordinate center point is converted to the absolute coordinate system through the transformation matrix. Therefore, in the actual use process, the laser radar needs to be calibrated in advance to obtain the transformation relationship between the local coordinate system with the laser radar as the coordinate center point and the absolute coordinate system.

[0066] In step S120, the acquisition device is used to acquire a first single frame of point cloud data of the real-time environment where the carrier is located in the local coordinate system.

[0067] In this step, the laser radar installed on the intelligent driving vehicle is used to collect point cloud data of the real-time environment of the vehicle frame by frame, that is, the first single-frame point cloud data. The laser radar scans the ground features identified by the real-time environment of the vehicle, such as traffic light poles, pedestrians and vehicles located above the ground, which are non-ground points, and lane line information located on the ground, which are ground points. The scanned data is recorded in the form of points, and each point cloud data includes three-dimensional coordinates, color information (R, G, B), and the intensity of the object's reflective surface.

[0068] In step S130, the coordinates of the first single-frame point cloud data are converted from the local coordinate system to the absolute coordinate system using the transformation relationship.

[0069] In this step, the coordinates of the first single-frame point cloud data are converted from the local coordinate system to the absolute coordinate system by utilizing the transformation relationship between the local coordinate system with the laser radar as the coordinate center point and the absolute coordinate system.

[0070] In step S140, a plane model is established using an algorithm.

[0071] In this step, the plane model is established by using an algorithm, which can be achieved by using the first single-frame point cloud data and using a random sampling consensus algorithm (RANSAC algorithm) to calculate mathematical model parameters of the first single-frame point cloud data to obtain the plane model.

[0072] The specific process includes: ① randomly select three points on the non-straight line from the first single-frame point cloud data, and calculate the mathematical model parameters that these three points meet; ② apply the plane model corresponding to the mathematical model parameters to all the first single-frame point cloud data, and separate the inner points and outer points of the plane model according to the threshold value of the point cloud data distance from the plane model; ③ randomly select three points on the non-straight line from the inner points, calculate the mathematical model parameters that these three points meet, and apply the plane model corresponding to the mathematical model parameters to all the inner points and re-separate the inner points and outer points; ④ repeat step ③ until there is no obvious change between the inner and outer points, and the distance residual of all outer points to the plane model is the smallest; ⑤ repeat steps ① to ④, and select an optimal mathematical model parameter as the final plane model.

[0073] In step S150, the first single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the acquisition device.

[0074] In this step, the first single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the laser radar. Specifically, the distance d1 of the first single-frame point cloud data from the plane model is calculated. When the distance d1 is less than the negative value of the elevation accuracy, the first single-frame point cloud data is marked as a noise point. When the distance d1 is greater than the elevation accuracy, the first single-frame point cloud data is marked as a non-ground point. When the distance d1 is greater than or equal to the negative value of the elevation accuracy and less than or equal to the elevation accuracy, the first single-frame point cloud data is marked as a ground point.

[0075] According to an embodiment of the present invention, the first single-frame point cloud data is used to calculate the mathematical model parameters of the first single-frame point cloud data using a random sampling consensus algorithm (RANSAC algorithm) to obtain a plane model. The first single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the laser radar. The first single-frame point cloud data is filtered and classified into ground points, non-ground points and noise points, thereby improving the anti-noise performance of the ground filtering method in processing the first single-frame point cloud data. By establishing a plane model, the first single-frame point cloud data collected by laser radars of different types and / or different installation methods is ground filtered, without the need to set multiple parameter thresholds, thereby improving the real-time and robustness of the ground filtering method in processing the first single-frame point cloud data.

[0076] Figure 2 It is a flowchart of a ground filtering method according to an embodiment of the present invention. Figure 3 FIG. 1 is a flow chart of a calibration method for a collection device according to an embodiment of the present invention. Figure 3 The calibration method of the acquisition device in the embodiment of the present invention is described in detail. Specifically, it includes the following steps:

[0077] In step S210, the acquisition device installed on the carrier is calibrated to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system.

[0078] In this step, according to Figure 3 The calibration method of the acquisition device shown calibrates the laser radar installed on the driving vehicle to obtain the external parameters of the laser radar to be calibrated.

[0079] Specifically, in step S310, a target normal equation group is established for the transformation relationship between the three-dimensional calibration field and the local coordinate system with the laser radar as the center point coordinates in the three-dimensional calibration field and the absolute coordinate system, and a plurality of high-reflective sheets with staggered heights and uniform distribution are set up in the three-dimensional calibration field as multiple control points.

[0080] The target normal equations for the transformation relationship between the local coordinate system and the absolute coordinate system are:

[0081]

[0082] in, is the second coordinate of the control point in the local coordinate system, is the first coordinate of the control point in the absolute coordinate system, It is the real-time dynamic posture information collected by the combined navigation equipment. is the rotation matrix of the integrated navigation device, is the coordinate of the integrated navigation device in the absolute coordinate system, and is the external parameter to be calibrated for the laser radar. Specifically, is the offset of the coordinates of the combined navigation device and the laser radar in the absolute coordinate system. is the rotation matrix of the lidar.

[0083] LiDAR external parameters to be calibrated

[0084]

[0085] in, are the unknown parameters that need to be solved.

[0086] In step S320, a real-time differential positioning method (RTK) is used to measure first coordinates of a plurality of control points in an absolute coordinate system.

[0087] In step S330, in the three-dimensional calibration field, according to the preset route, the radar device is used to collect the third single-frame point cloud data of the real-time environment of the driving vehicle in the local coordinate system with the radar device as the coordinate center point, and the integrated navigation device carried by the driving vehicle is used to collect real-time dynamic posture information. The real-time dynamic posture information may include the rotation matrix of the integrated navigation device and the coordinates of the integrated navigation device in the absolute coordinate system.

[0088] In step S340, VeloView is used to play the original data packet of the third single-frame point cloud data collected by the laser radar, and the multiple control points are manually selected to obtain the second coordinates of the multiple control points in the local coordinate system with the laser radar as the center point coordinate. The first coordinates of the multiple control points in the absolute coordinate system and the second coordinates of the multiple control points in the local coordinate system form a one-to-one corresponding coordinate point pair.

[0089] In step S350, based on the first coordinates, the second coordinates, and the posture information of the multiple control points, a least squares algorithm is used to solve the target normal equation group to obtain the external parameters to be calibrated of the laser radar.

[0090] In step S360, based on the laser radar's extrinsic parameters to be calibrated and the target normal equation group, i.e., formula (1), the transformation relationship between the local coordinate system with the laser radar as the center point coordinate and the absolute coordinate system is obtained. The normal equation group of the transformation relationship between the local coordinate system with the laser radar as the center point coordinate and the absolute coordinate system is:

[0091]

[0092] in, is the coordinate of the point cloud data in the local coordinate system with the laser radar as the center point coordinate, is the coordinate of the point cloud data in the absolute coordinate system, It is the real-time dynamic posture information collected by the combined navigation equipment. is the rotation matrix of the integrated navigation device, is the coordinate of the integrated navigation device in the absolute coordinate system, It is the external parameter of the laser radar calibrated in step S350.

[0093] In step S220, the acquisition device is used to acquire a first single frame of point cloud data of the real-time environment where the carrier is located in the local coordinate system.

[0094] Step 220 and Figure 1 The process is consistent with step S120 in the embodiment, and will not be described in detail here.

[0095] In step S230, the coordinates of the first single-frame point cloud data are converted from the local coordinate system to the absolute coordinate system using the transformation relationship.

[0096] In this step, the coordinates of the first single frame point cloud data are converted from the local coordinate system to the absolute coordinate system using formula (3).

[0097] In step S240, a first height of the acquisition device installed on the carrier from the ground is obtained.

[0098] In this step, the laser radar can be installed on the top of the driving vehicle to measure the first height H1 of the laser radar from the ground after the laser radar is set up and installed on the driving vehicle.

[0099] In step S250, the coordinates of the acquisition device are converted from the local coordinate system to the absolute coordinate system using the transformation relationship.

[0100] In this step, the coordinates (0,0,0) of the laser radar are converted from the local coordinate system to the absolute coordinate system using formula (3) to obtain the position coordinates of the laser radar in the real world:

[0101] In step S260, a second height of the ground in the absolute coordinate system is calculated according to the first height and the coordinates of the acquisition device in the absolute coordinate system.

[0102] In this step, according to the first height H1 and the coordinates of the laser radar in the absolute coordinate system Calculate the second height of the ground in the absolute coordinate system

[0103] In step S270, the first single-frame point cloud data is filtered according to the second height and the first threshold to obtain second single-frame point cloud data.

[0104] In this step, the first single-frame point cloud data is filtered according to the second height H2 and the first threshold L to obtain the second single-frame point cloud data. The first threshold L can be set manually, for example, 0.5 meters. The first single-frame point cloud data is filtered to obtain the second single-frame point cloud data with a height range of The point cloud data within the frame is obtained to obtain the second single-frame point cloud data.

[0105] In step S280, a plane model is established using an algorithm.

[0106] In this step, the plane model is established by using an algorithm, which can be achieved by using the second single-frame point cloud data and using a random sampling consensus algorithm (RANSAC algorithm) to calculate mathematical model parameters of the second single-frame point cloud data to obtain the plane model.

[0107] The specific process includes: ① randomly select three points on the non-straight line from the second single-frame point cloud data, and calculate the mathematical model parameters that these three points meet; ② apply the plane model corresponding to the mathematical model parameters to all the second single-frame point cloud data, and separate the inner points and outer points of the plane model according to the threshold value of the point cloud data distance from the plane model; ③ randomly select three points on the non-straight line from the inner points, calculate the mathematical model parameters that these three points meet, and apply the plane model corresponding to the mathematical model parameters to all the inner points and re-separate the inner points and outer points; ④ repeat step ③ until there is no obvious change between the inner and outer points, and the distance residual of all outer points to the plane model is the smallest; ⑤ repeat steps ① to ④, and select an optimal mathematical model parameter as the final plane model.

[0108] In step S290, the second single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the acquisition device.

[0109] In this step, the second single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the laser radar. Specifically, the distance d2 of the second single-frame point cloud data from the plane model is calculated. When the distance d2 is less than the negative value of the elevation accuracy, the second single-frame point cloud data is marked as a noise point. When the distance d2 is greater than the elevation accuracy, the second single-frame point cloud data is marked as a non-ground point. When the distance d2 is greater than or equal to the negative value of the elevation accuracy and less than or equal to the elevation accuracy, the second single-frame point cloud data is marked as a ground point.

[0110] Repeat steps S220 to S290 to process each frame of point cloud data in real time, implement ground filtering to process the point cloud data of the real-time surrounding environment of the driving vehicle collected by the lidar, and then provide decision-making for intelligent driving.

[0111] According to an embodiment of the present invention, after the laser radar is set up and installed on the driving vehicle, the first height of the laser radar from the ground is measured. According to the first height and the coordinates of the laser radar in the absolute coordinate system, the second height of the ground in the absolute coordinate system is calculated. According to the second height and the first threshold, the first single-frame point cloud data is filtered to obtain the second single-frame point cloud data. Using the second single-frame point cloud data, the mathematical model parameters of the second single-frame point cloud data are calculated using a random sampling consensus algorithm (RANSAC algorithm) to obtain a plane model. The second single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the laser radar. Filtering the first single-frame point cloud data to obtain the second single-frame point cloud data reduces the amount of calculation for subsequent filtering of the second single-frame point cloud data according to the plane model and the elevation accuracy of the laser radar, and further improves the real-time performance of the ground filtering method in processing the first single-frame point cloud data.

[0112] In addition, a target normal equation group of the transformation relationship between the local coordinate system with the laser radar as the center point coordinate and the absolute coordinate system in the three-dimensional calibration field is established. The first coordinates of multiple control points in the absolute coordinate system in the three-dimensional calibration field and the second coordinates of multiple control points in the local coordinate system form a one-to-one corresponding coordinate point pair. According to the first coordinates, the second coordinates, and the posture information of the multiple control points, the least squares algorithm is used to solve the target normal equation group to obtain the extrinsic parameters of the laser radar to be calibrated. Based on the calibrated extrinsic parameters of the laser radar and the target normal equation group, the transformation relationship between the local coordinate system with the laser radar as the center point coordinate and the absolute coordinate system is obtained. The external parameter calibration method of the laser radar in the embodiment of the present invention is used to calibrate the external parameters of the laser radar to be calibrated, which improves the accuracy of the external parameters of the calibrated laser radar, thereby improving the accuracy of using the vehicle-mounted laser radar to scan the real-time environment around the driving vehicle.

[0113] Figure 4 Schematic diagram of the structure of a ground filter device according to an embodiment of the present invention. The ground filter device according to the embodiment of the present application is applied to intelligent driving technology. Figure 4As shown, the ground filtering device includes: a calibration unit 410 , a collection unit 420 , a first coordinate transformation unit 430 , a building unit 440 and a first filtering unit 450 .

[0114] The calibration unit 410 is configured to calibrate the laser radar installed on the driving vehicle and obtain the transformation relationship between the local coordinate system with the laser radar as the center point coordinate and the absolute coordinate system.

[0115] The acquisition unit 420 is configured to use the laser radar to acquire the first single frame point cloud data of the real-time environment where the driving vehicle is located in the local coordinate system.

[0116] The first coordinate transformation unit 430 is configured to transform the coordinates of the first single-frame point cloud data from the local coordinate system to the absolute coordinate system by using a transformation relationship.

[0117] The establishing unit 440 is configured to execute the establishment of the plane model by using an algorithm, which may be to use the first single frame point cloud data and adopt a random sampling consensus algorithm (RANSAC algorithm) to calculate the mathematical model parameters of the first single frame point cloud data to obtain the plane model.

[0118] The first filtering unit 450 is configured to filter the first single frame point cloud data according to the plane model and the elevation accuracy of the laser radar.

[0119] Figure 5 FIG. 1 is a schematic diagram of the structure of a ground filtering device according to an embodiment of the present invention. Figure 5 As shown, the ground filtering device includes: a calibration unit 510, a collection unit 520, a first coordinate transformation unit 530, an acquisition unit 540, a second coordinate transformation unit 550, a calculation unit 560, a second filtering unit 570, an establishment unit 580 and a first filtering unit 590.

[0120] The calibration unit 510 is configured to calibrate the laser radar installed on the driving vehicle and obtain the transformation relationship between the local coordinate system with the laser radar as the center point coordinate and the absolute coordinate system.

[0121] Specifically, a target normal equation group of the transformation relationship between the local coordinate system and the absolute coordinate system in the three-dimensional calibration field is established, and multiple control points are set in the three-dimensional calibration field. The first coordinates of multiple control points in the absolute coordinate system are measured using a real-time differential positioning method (RTK). In the three-dimensional calibration field, according to the preset route, the third single-frame point cloud data of the real-time environment where the driving vehicle is located in the local coordinate system is collected using a laser radar. Real-time dynamic posture information is collected using a combined navigation device. The multiple control points are selected from the third single-frame point cloud data to obtain the second coordinates of the multiple control points in the local coordinate system. According to the first coordinates, second coordinates, and posture information of the multiple control points, the least squares algorithm is used to solve the target normal equation group to obtain the extrinsic parameters to be calibrated of the laser radar. Based on the extrinsic parameters to be calibrated of the laser radar and the target normal equation group, the transformation relationship between the local coordinate system and the absolute coordinate system is obtained.

[0122] The acquisition unit 520 is configured to use the laser radar to acquire the first single frame point cloud data of the real-time environment where the driving vehicle is located in the local coordinate system.

[0123] The first coordinate transformation unit 530 is configured to transform the coordinates of the first single-frame point cloud data from the local coordinate system to the absolute coordinate system by using a transformation relationship.

[0124] The acquisition unit 540 is configured to acquire a first height of a laser radar installed on the driving vehicle from the ground.

[0125] The second coordinate transformation unit 550 is configured to transform the coordinates of the laser radar from the local coordinate system to the absolute coordinate system using a transformation relationship.

[0126] The calculation unit 560 is configured to calculate a second height of the ground in the absolute coordinate system based on the first height and the coordinates of the laser radar in the absolute coordinate system.

[0127] The second filtering unit 570 is configured to filter the first single-frame point cloud data according to the second height and the first threshold to obtain second single-frame point cloud data.

[0128] The establishing unit 580 is configured to execute the establishment of the plane model by using an algorithm, which may be to use the second single-frame point cloud data and adopt a random sampling consensus algorithm (RANSAC algorithm) to calculate the mathematical model parameters of the second single-frame point cloud data to obtain the plane model.

[0129] The first filtering unit 590 is configured to filter the first single frame point cloud data according to the plane model and the elevation accuracy of the acquisition device.

[0130] Figure 6 A structural diagram of a ground filtering control device according to an embodiment of the present invention. Figure 6The device shown is only an example and should not constitute any limitation to the functions and scope of use of the embodiments of the present invention.

[0131] refer to Figure 6 , the device includes a processor 610, a memory 620 and an input-output device 630 connected by a bus. The memory 620 includes a read-only memory (ROM) and a random access memory (RAM). The memory 620 stores various computer instructions and data required to perform system functions. The processor 610 reads various computer instructions from the memory 620 to perform various appropriate actions and processes. The input-output device includes an input part such as a keyboard, a mouse, etc.; an output part such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage part such as a hard disk; and a communication part such as a network interface card such as a LAN card, a modem, etc. The memory 620 also stores computer instructions to complete the operations specified in the ground filtering method of the embodiment of the present invention.

[0132] Accordingly, an embodiment of the present invention provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed, the operations specified by the above-mentioned ground filtering method are implemented.

[0133] Correspondingly, an embodiment of the present invention further provides a computer program product, including a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a mobile terminal, the mobile terminal executes the steps of the above-mentioned ground filtering method.

[0134] The flowcharts and block diagrams in the accompanying drawings illustrate the possible system frameworks, functions and operations of the systems, methods and devices of the embodiments of the present invention. The boxes on the flowcharts and block diagrams can represent a module, a program segment or just a piece of code, and the modules, program segments and codes are all executable instructions for implementing the specified logical functions. It should also be noted that the executable instructions for implementing the specified logical functions can be recombined to generate new modules and program segments. Therefore, the boxes and the order of the boxes in the accompanying drawings are only used to better illustrate the process and steps of the embodiments, and should not be used as a limitation on the invention itself.

[0135] The above descriptions are only some embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A ground filtering method, characterized in that: include: Calibrate the acquisition device installed on the carrier to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system; Using the acquisition device to acquire first single-frame point cloud data of the real-time environment where the carrier is located in the local coordinate system; Converting the coordinates of the first single-frame point cloud data from the local coordinate system to the absolute coordinate system using the transformation relationship; Establishing a plane model using an algorithm, wherein the second single-frame point cloud data is used to calculate mathematical model parameters of the second single-frame point cloud data using a random sampling consensus algorithm to obtain the plane model, wherein the first single-frame point cloud data is filtered according to a second height of the ground in the absolute coordinate system and a first threshold to obtain the second single-frame point cloud data; as well as The first single-frame point cloud data is filtered according to the plane model and the elevation accuracy of the acquisition device, wherein the distance of the second single-frame point cloud data from the plane model is calculated, when the distance is less than the negative value of the elevation accuracy, the second single-frame point cloud data is marked as a noise point, when the distance is greater than the elevation accuracy, the second single-frame point cloud data is marked as a non-ground point, and when the distance is greater than or equal to the negative value of the elevation accuracy and less than or equal to the elevation accuracy, the second single-frame point cloud data is marked as a ground point.

2. The ground filtering method according to claim 1, characterized in that: The method of using an algorithm to establish a plane model includes: Acquire a first height of the acquisition device mounted on the carrier from the ground; The coordinates of the acquisition device are converted from the local coordinate system to the absolute coordinate system using the transformation relationship.

3. The ground filtering method according to claim 2, characterized in that: Before the plane model is established by the algorithm, the following steps are also included: Calculating a second height of the ground in the absolute coordinate system according to the first height and the coordinates of the acquisition device in the absolute coordinate system; The first single-frame point cloud data is filtered according to the second height and the first threshold to obtain second single-frame point cloud data.

4. The ground filtering method according to claim 1, characterized in that: The calibrating of the acquisition device mounted on the carrier to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system includes: Establishing a target normal equation group of the transformation relationship between the local coordinate system and the absolute coordinate system in the three-dimensional calibration field, and setting a plurality of control points in the three-dimensional calibration field; Measuring first coordinates of the plurality of control points in the absolute coordinate system using a real-time differential positioning method; In the three-dimensional calibration field, according to a preset route, using the acquisition device to acquire a third single-frame point cloud data of the real-time environment where the carrier is located in the local coordinate system; and Use integrated navigation equipment to collect real-time dynamic posture information.

5. The ground filtering method according to claim 4, characterized in that: The calibrating of the acquisition device mounted on the carrier to obtain the transformation relationship between the local coordinate system with the acquisition device as the center point coordinate and the absolute coordinate system further includes: Selecting the plurality of control points from the third single-frame point cloud data to obtain second coordinates of the plurality of control points in the local coordinate system; Solving the target normal equations using a least squares algorithm based on the first coordinates, the second coordinates, and the pose information of the plurality of control points to obtain the extrinsic parameters to be calibrated for the acquisition device; and Based on the extrinsic parameters to be calibrated of the acquisition device and the target normal equation group, the transformation relationship between the local coordinate system and the absolute coordinate system is obtained.

6. The ground filtering method according to claim 5, characterized in that: The target normal equations for the transformation relationship between the local coordinate system and the absolute coordinate system are: in, is the second coordinate of the control point in the local coordinate system, is the first coordinate of the control point in the absolute coordinate system, is the real-time dynamic posture information collected by the integrated navigation device, is the rotation matrix of the integrated navigation device, is the coordinate of the integrated navigation device in the absolute coordinate system, It is the external parameter to be calibrated of the acquisition device.

7. A ground filtering device, characterized in that: include: a calibration unit configured to calibrate the acquisition device mounted on the carrier and obtain a transformation relationship between a local coordinate system with the acquisition device as the center point coordinate and an absolute coordinate system; An acquisition unit is configured to acquire a first single-frame point cloud data of the real-time environment where the carrier is located in the local coordinate system using the acquisition device; A first coordinate transformation unit is configured to transform the coordinates of the first single-frame point cloud data from the local coordinate system to the absolute coordinate system using the transformation relationship; an establishing unit configured to execute establishing a plane model using an algorithm, wherein a second single-frame point cloud data is used to calculate mathematical model parameters of the second single-frame point cloud data using a random sampling consensus algorithm to obtain the plane model, wherein the first single-frame point cloud data is filtered according to a second height of the ground in the absolute coordinate system and a first threshold to obtain the second single-frame point cloud data; as well as The first filtering unit is configured to perform filtering of the first single-frame point cloud data according to the plane model and the elevation accuracy of the acquisition device, wherein the distance of the second single-frame point cloud data from the plane model is calculated, and when the distance is less than the negative value of the elevation accuracy, the second single-frame point cloud data is marked as a noise point; when the distance is greater than the elevation accuracy, the second single-frame point cloud data is marked as a non-ground point; when the distance is greater than or equal to the negative value of the elevation accuracy and less than or equal to the elevation accuracy, the second single-frame point cloud data is marked as a ground point.

8. The ground filtering device according to claim 7, characterized in that: Before the plane model is established by the algorithm, the ground filtering device further includes: An acquiring unit is configured to acquire a first height of the acquisition device mounted on the carrier from the ground; The second coordinate transformation unit is configured to transform the coordinates of the acquisition device from the local coordinate system to the absolute coordinate system by using the transformation relationship.

9. The ground filtering device according to claim 8, characterized in that: Before the plane model is established by the algorithm, the ground filtering device further includes: A calculation unit is configured to calculate a second height of the ground in the absolute coordinate system according to the first height and the coordinates of the acquisition device in the absolute coordinate system; The second filtering unit is configured to filter the first single-frame point cloud data according to the second height and the first threshold to obtain second single-frame point cloud data.

10. A ground filter control device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the ground filtering method according to any one of claims 1 to 6.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and when the computer instructions are executed, the ground filtering method according to any one of claims 1 to 6 is implemented.

12. A computer program product, comprising a computer program, wherein the computer program comprises program instructions, and when the program instructions are executed by a mobile terminal, the mobile terminal is caused to perform the steps of the terrestrial filtering method according to any one of claims 1 to 6.

Images