A laser radar internal parameter calibration method and device

Abstract

The present invention discloses a laser radar internal parameter calibration method and device, including a data acquisition step: sampling point cloud data of a calibration target in different postures at different calibration test points in a calibration environment; a data processing step: performing point cloud segmentation on the sampled point cloud data, extracting calibration marker data and performing marker vertex fitting; an internal parameter optimization step: constructing an internal parameter optimization model using the fitted vertex coordinates, performing internal parameter solution, and obtaining optimal parameters; a performance verification step: obtaining the distance between markers on the calibration target based on the solved optimal parameters, and verifying the performance after calibration. The present invention constructs a universal laser radar internal parameter calibration model for mechanical, semi-solid and solid-state laser radars through a method for dynamically acquiring and processing laser radar calibration data; designs a new laser radar calibration device, which does not require measuring the distance between the laser radar and the target, and uses the calibration method disclosed in the present invention to improve calibration efficiency and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of laser radar calibration, and in particular to a laser radar internal parameter calibration method and device. Background Art

[0002] With the rapid development of autonomous vehicles, intelligent robots, and other fields, the demand for high-precision three-dimensional environmental perception is becoming increasingly urgent. Due to its high environmental adaptability and high-precision ranging capabilities, LiDAR (LiDAR) has become the primary sensing device in these fields. However, due to factors such as LiDAR's complex optical path design, manufacturing, and installation processes, if its internal parameters are not properly calibrated, the output 3D measurement information may deviate significantly from the actual situation, seriously affecting the accuracy of environmental perception.

[0003] LiDAR intrinsic calibration relies heavily on the design of calibration targets. Currently, most use planar targets to obtain calibration point cloud information. However, this method requires precise external measurements, is subject to significant human influence, and is extremely time-consuming and labor-intensive. In recent years, many manufacturers have proposed new calibration target designs. For example, patent [201810982381.6] describes a new planar calibration target with a photosensitive plate. This plate's automatic sensing capability is used to obtain calibration point cloud data. However, this calibration target is extremely expensive to manufacture and requires precise measurement of the target's absolute coordinates relative to the LiDAR. Patent [201710515747.4] describes new calibration targets using cones and spheres. However, the measurement accuracy of these targets' installation positions is extremely demanding. Furthermore, existing LiDAR intrinsic calibration models are only applicable to mechanical LiDARs and are not suitable for semi-solid-state and solid-state LiDARs. In summary, these technical and methodological limitations have led to high LiDAR production costs and difficulty ensuring consistent measurement accuracy. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the existing technology. To achieve the above purpose, a laser radar internal parameter calibration method and device are adopted to solve the problems raised in the above background technology.

[0005] A laser radar internal parameter calibration method, comprising:

[0006] Step S1, data acquisition step: sampling point cloud data of the calibration target in different postures at different calibration test points in the calibration environment;

[0007] Step S2, data processing step: performing point cloud segmentation on the sampled point cloud data, extracting calibration marker data and performing marker vertex fitting;

[0008] Step S3, internal parameter optimization step: construct an internal parameter optimization model using the fitted vertex coordinates, perform internal parameter calculation, and obtain the optimal parameters;

[0009] Step S4, performance verification step: based on the calculated optimal parameters, the distance between the markers on the verification target is obtained to verify the performance after calibration.

[0010] As a further solution of the present invention: the data acquisition step in step S1 specifically includes:

[0011] Step S11: configuring the initial internal parameters of the laser radar according to the design parameters;

[0012] Step S12: Install the laser radar on a movable calibration platform;

[0013] Step S13: Control the calibration platform to move to each calibration test point in sequence according to the calibration test point sequence;

[0014] Step S14: At each calibration test point, adjust the angle between the laser radar and the horizontal plane by adjusting the angle bracket, sample the calibration targets at different positions in the calibration environment, and obtain point cloud data under the current parameter configuration;

[0015] Step S15: Complete data sampling at all calibration test points and store the acquired data.

[0016] As a further solution of the present invention: the data processing step in step S2 specifically includes:

[0017] Step S21, read the point cloud data stored in step S2, first perform calibration target plane segmentation on each frame of point cloud, use the RANSAC algorithm to extract all planes in the current point cloud, and segment the calibration target plate and the cone calibration object mounted on it from the point cloud;

[0018] Step S22, using a clustering algorithm to further segment the point cloud data segmented in the previous step, and extract the markers installed on the verification target plate;

[0019] In step S23, based on the surface point cloud data of the marker extracted in the previous step, the corresponding conical surface equation parameters are fitted using the least squares method. The least squares loss function is defined as follows:

[0020]

[0021] In the above formula, (x0, y0, z0) represents the vertex of the cone, (m, n, p) represents the direction vector of the cone axis, and θ is the angle between the cone generatrix and the axis;

[0022] The extracted conical surface data points (x i ,y i,z i ) into the above equation and minimize the loss function through iterative optimization to solve the value of the parameters (x0, y0, z0, m, n, p, θ);

[0023] In step S24, the coordinates of the cone vertices of all markers on the same calibration target are solved using step S23, and then the average error between the distances between the coordinates of the cone vertices of different cones and the true value is calculated, that is:

[0024]

[0025] In the above formula, (x i ,y i ,z i ), i = 1, 2, 3, 4 are the coordinates of the apexes of four different cones obtained by fitting on the same target, and d0 is the actual distance between them.

[0026] As a further solution of the present invention, the internal reference optimization step in step S3 specifically includes:

[0027] Step S31: Establish a laser radar internal parameter calibration model, where the internal parameter is in the form of a calibration matrix:

[0028]

[0029] In the above formula, R represents the 3D rotation matrix, t is the translation vector, and s is the scale factor. For any measurement value x, it can be converted into another 3D Cartesian coordinate through the H matrix:

[0030] x′=Hx=sRx+t;

[0031] Step 32: construct an internal parameter optimization model, assuming that the sampling point cloud of a specific calibration target plate calibration object under a specific internal parameter configuration is T represents the number of scanning postures at the same scanning position. The optimization loss can be defined as:

[0032]

[0033] In the above formula, N is the number of sampling positions, M is the number of calibration target plates for a single sampling, and Δd nmt That is, the average error between the distances between the coordinates of the cone apexes of different cones on the same calibration target obtained in step S24 and the true value; since J implicitly expresses the error relationship between the distances between the intrinsic parameter matrix H and the feature points on the calibration target and the true value, the optimal value of H is estimated by optimizing J;

[0034] Step S33: Based on a heuristic optimization algorithm, an optimal estimate of the internal parameter matrix is calculated through iterative optimization.

[0035] As a further solution of the present invention: the performance verification step in step S4 specifically includes:

[0036] Step S41: Based on the optimized internal parameters, the laser radar is configured to scan two calibration target plates that are not within the original calibration range to obtain corresponding sampling point clouds;

[0037] Step S42, then using the method of step S24, obtain the errors between the coordinates of the cone apexes of each cone;

[0038] Step S43: If the error is less than the threshold d TH , the calibration process is terminated and the calibration parameters are output;

[0039] Step S44: If the error is greater than the threshold d TH , then re-collect the calibration data and re-calibrate the parameters.

[0040] Another technical solution is a device for a laser radar internal parameter calibration method, which adopts a laser radar internal parameter calibration method as described in any one of the above items, and the device includes:

[0041] Calibration target, used for calibration in combination with cone markers;

[0042] Verification targets, used to obtain data at longer distances than the calibration targets;

[0043] Calibration platform, used to provide a lidar calibration test platform;

[0044] Communication module, used to transmit lidar data and calibration processing results of the computing system;

[0045] A computing system, configured to send motion control instructions to the calibration platform through a communication module according to the positions of calibration test points;

[0046] The storage server is used to receive the lidar data and lidar calibration results sent by the computing system, and store the lidar data and lidar calibration results in the form of files.

[0047] As a further solution of the present invention: the calibration platform includes a laser radar, an angle adjustment bracket, a mobile platform, and a calibration platform controller.

[0048] As a further solution of the present invention: the communication module includes a wireless serial port, a wireless network port, a first network port, and a second network port.

[0049] Compared with the prior art, the present invention has the following technical effects:

[0050] By adopting the above technical solution, the automation level of the calibration process is improved through the dynamic acquisition and processing method of lidar calibration data; secondly, a universal lidar internal parameter calibration model is constructed, which can be applied to mechanical, semi-solid and solid-state lidars; finally, a new lidar calibration device is designed, which does not require measuring the distance between the lidar and the target. The method of the present invention is used to improve the calibration efficiency and calibration accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings:

[0052] Figure 1 A schematic diagram of the steps of the calibration method disclosed in the present application;

[0053] Figure 2 This is a block diagram of the laser radar internal parameter calibration device according to the embodiment disclosed in this application;

[0054] Figure 3 This is a diagram showing an embodiment of a laser radar internal parameter calibration device according to an embodiment disclosed in this application;

[0055] Figure 4 This is a schematic diagram of the structure of the calibration platform of the embodiment disclosed in this application;

[0056] Figure 5 A schematic diagram of a calibration target according to an embodiment disclosed in this application;

[0057] Figure 6 A schematic diagram of a calibration target according to an embodiment disclosed in this application;

[0058] Figure 7 This is a schematic diagram of a cone marker according to an embodiment disclosed in this application. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0060] Please refer to Figure 1 In an embodiment of the present invention, a laser radar internal parameter calibration method includes the following steps:

[0061] Step S1, data acquisition step: sampling point cloud data of the calibration target in different postures at different calibration test points in the calibration environment, specifically including:

[0062] Step S11: configuring the initial internal parameters of the laser radar according to the design parameters;

[0063] Step S12: Install the laser radar on a movable calibration platform;

[0064] Step S13: Control the calibration platform to move to each calibration test point in sequence according to the calibration test point sequence;

[0065] Step S14: At each calibration test point, adjust the angle between the laser radar and the horizontal plane by adjusting the angle bracket, sample the calibration targets at different positions in the calibration environment, and obtain point cloud data under the current parameter configuration;

[0066] Step S15: Complete data sampling at all calibration test points and store the acquired data.

[0067] After collecting all calibration test point data, enter the data processing step of step S2;

[0068] Step S2, data processing step: performing point cloud segmentation on the sampled point cloud data, extracting calibration marker data and performing marker vertex fitting, specifically including:

[0069] Step S21: Read the point cloud data stored in step S2, first perform calibration target plane segmentation on each frame of point cloud, use the RANSAC algorithm to extract all planes in the current point cloud, and segment the calibration target plate and the cone calibration object mounted on it from the point cloud. In this embodiment, the calibration object is a cone.

[0070] Step S22, using a clustering algorithm to further segment the point cloud data segmented in the previous step, and extract the markers installed on the verification target plate;

[0071] In step S23, based on the surface point cloud data of the marker extracted in the previous step, the corresponding conical surface equation parameters are fitted using the least squares method. The least squares loss function is defined as follows:

[0072]

[0073] In the above formula, (x0, y0, z0) represents the vertex of the cone, (m, n, p) represents the direction vector of the cone axis, and θ is the angle between the cone generatrix and the axis;

[0074] The extracted conical surface data points (x i ,y i ,z i ) into the above equation and minimize the loss function through iterative optimization to solve the value of the parameters (x0, y0, z0, m, n, p, θ);

[0075] In step S24, the coordinates of the cone vertices of all markers on the same calibration target are solved using step S23, and then the average error between the distances between the coordinates of the cone vertices of different cones and the true value is calculated, that is:

[0076]

[0077] In the above formula, (x i ,y i ,z i ), i = 1, 2, 3, 4 are the coordinates of the apexes of four different cones obtained by fitting on the same target, and d0 is the actual distance between them.

[0078] Step S3, internal parameter optimization step: using the fitted vertex coordinates to build an internal parameter optimization model, perform internal parameter calculation, and obtain the optimal parameters, specifically including:

[0079] Step S31: Establish a laser radar internal parameter calibration model, where the internal parameter is in the form of a calibration matrix:

[0080]

[0081] In the above formula, R represents the 3D rotation matrix, t is the translation vector, and s is the scale factor. For any measurement value x, it can be converted into another 3D Cartesian coordinate through the H matrix:

[0082] x′=Hx=sRx+t;

[0083] Step 32: construct an internal parameter optimization model, assuming that the sampling point cloud of a specific calibration target plate calibration object under a specific internal parameter configuration is T represents the number of scanning postures at the same scanning position. The optimization loss can be defined as:

[0084]

[0085] In the above formula, N is the number of sampling positions, M is the number of calibration target plates for a single sampling, and Δd nmt That is, the average error between the distances between the coordinates of the cone apexes of different cones on the same calibration target obtained in step S24 and the true value; since J implicitly expresses the error relationship between the distances between the intrinsic parameter matrix H and the feature points on the calibration target and the true value, the optimal value of H is estimated by optimizing J;

[0086] Step S33: Based on a heuristic optimization algorithm, an optimal estimate of the internal parameter matrix is calculated through iterative optimization.

[0087] Step S4, performance verification step: obtaining the distance between markers on the verification target based on the calculated optimal parameters and verifying the performance after calibration, specifically including:

[0088] Step S41: Based on the optimized internal parameters, the laser radar is configured to scan two calibration target plates that are not within the original calibration range to obtain corresponding sampling point clouds;

[0089] Step S42, then using the method of step S24, obtain the errors between the coordinates of the cone apexes of each cone;

[0090] Step S43: If the error is less than the threshold d TH (0.5mm), the calibration process is terminated and the calibration parameters are output;

[0091] Step S44: If the error is greater than the threshold d TH (0.5mm), re-collect the calibration data and re-calibrate the parameters.

[0092] Another technical solution is a device for a laser radar internal parameter calibration method, which adopts a laser radar internal parameter calibration method as described in any one of the above items, and the device includes:

[0093] Calibration target, used for calibration in combination with cone markers;

[0094] Verification targets, used to obtain data at longer distances than the calibration targets;

[0095] Calibration platform, used to provide a lidar calibration test platform;

[0096] Communication module, used to transmit lidar data and calibration processing results of the computing system;

[0097] A computing system, configured to send motion control instructions to the calibration platform through a communication module according to the positions of calibration test points;

[0098] The storage server is used to receive the lidar data and lidar calibration results sent by the computing system, and store the lidar data and lidar calibration results in the form of files.

[0099] In this embodiment, the calibration platform includes a laser radar, an angle adjustment bracket, a mobile platform, and a calibration platform controller.

[0100] In this embodiment, the communication module includes a wireless serial port, a wireless network port, a first network port, and a second network port.

[0101] In a specific embodiment, Figure 3 As shown, the calibration environment (208) is a flat indoor space for placing calibration targets, verification targets, calibration platforms, communication modules, computing systems, and storage servers. Circular calibration test points (205) are drawn on the ground of the calibration environment. The path formed by the discrete circular calibration test points serves as an indication path for the movement of the calibration platform.

[0102] like Figure 2 、 Figure 3 and Figure 4 As shown, the calibration platform (202) is composed of a laser radar, an angle adjustment bracket, a mobile platform, and a calibration platform controller. The calibration platform controller can drive and control the movement of the mobile platform through control instructions received through the wireless serial port 1. The laser radar and the angle adjustment bracket are installed on the mobile platform. The angle adjustment bracket can adjust the angle between the laser radar and the horizontal plane. The laser radar and the calibration platform controller are connected via a data transmission cable.

[0103] like Figure 3 and Figure 5 As shown, the calibration target (201) is placed in front of and on both sides of the calibration platform. The plane target plate is designed in combination with the cone. The area of the plane target plate is 1500×1500 mm2. Four cones are fixed on the calibration target plate with a spacing of d0=750 mm and an installation position accuracy of ±0.1 mm. The cone (such as Figure 7 The bottom circle diameter d is 300mm and the height h is 150mm.

[0104] like Figure 3 and Figure 6 As shown, the calibration target (207) is placed in front of the calibration platform. Similar to the calibration target, its plane target plate area is 2000×2000 mm2 in order to obtain data at a longer distance. Five cones are fixed on the calibration target plate with a spacing of d1=1000 mm and an installation position accuracy of ±0.1 mm. Figure 7 The bottom circle diameter of the container (as shown) is 400 mm and the height is 150 mm.

[0105] like Figure 2 and Figure 3 As shown, the communication module (203) has interfaces such as a wireless serial port 2, a wireless second network port 2, a first network port 1, and a second network port 2. The network port 3 of the computing system is connected to the first network port 1 of the communication module to transmit motion control instructions and laser radar data. The wireless serial port 2 of the communication module is connected to the wireless serial port 1 of the calibration platform controller to transmit motion control instructions, and the wireless second network port 2 of the communication module is connected to the wireless first network port 1 of the calibration platform controller to transmit laser radar data. The network port 4 of the storage server is connected to the second network port 2 of the communication module to transmit laser radar data and the calibration processing results of the computing system.

[0106] like Figure 2 and Figure 3As shown, the computing system (204) sends a motion control instruction to the calibration platform controller through the wireless serial port 2 of the communication module according to the position of the calibration test point, thereby controlling the calibration platform to move to the calibration test point. After the calibration platform reaches the calibration test point, the laser radar collects data of the calibration target and the verification target. The calibration platform controller transmits the data to the computing system through the wireless first network port 1 through the wireless second network port 2 and the first network port 1 of the communication module. The computing system calculates the calibration result of the laser radar based on the sampled laser radar data. The computing system transmits the laser radar data and the calibration result to the data storage server through the first network port 1 and the second network port 2 of the communication module.

[0107] like Figure 2 and Figure 3 As shown, the storage server (206) is connected to the second network port 2 of the communication module via the network port 4, and mainly receives the laser radar data and laser radar calibration results sent by the computing system, and stores the laser radar data and laser radar calibration results in the form of files in the internal storage unit. The laser radar data and laser radar calibration results can be displayed by the display unit of the storage server.

[0108] Although the embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents, and all should be included within the scope of protection of the present invention.

Claims

1. A laser radar internal parameter calibration method, characterized in that: The specific steps include: Step S1, data acquisition step: sampling point cloud data of the calibration target in different postures at different calibration test points in the calibration environment; Step S2, data processing step: performing point cloud segmentation on the sampled point cloud data, extracting calibration marker data and performing marker vertex fitting, which specifically includes: Step S21, read the point cloud data stored in step S2, first perform calibration target plane segmentation on each frame of point cloud, use the RANSAC algorithm to extract all planes in the current point cloud, and segment the calibration target plate and the cone calibration object mounted on it from the point cloud; Step S22, using a clustering algorithm to further segment the point cloud data segmented in the previous step, and extract the markers installed on the verification target plate; In step S23, based on the surface point cloud data of the marker extracted in the previous step, the corresponding conical surface equation parameters are fitted using the least squares method. The least squares loss function is defined as follows: In the above formula, ( x 0, y 0, z 0) represents the vertex of the cone, ( m , n , p ) represents the direction vector of the cone axis, θ is the angle between the cone generatrix and the axis; The extracted conical surface data points ( x i , y i , z i ) is substituted into the above equation and the loss function is minimized through iterative optimization to solve the parameter ( x 0, y 0, z 0, m , n , p , θ ) value; In step S24, the coordinates of the cone vertices of all markers on the same calibration target are solved using step S23, and then the average error between the distances between the coordinates of the cone vertices of different cones and the true value is calculated, that is: In the above formula, ( x i , y i , z i ), i =1, 2, 3, 4 are the coordinates of the tops of four different cones obtained by fitting on the same target. d 0 is the actual distance between them; Step S3, internal parameter optimization step: construct an internal parameter optimization model using the fitted vertex coordinates, perform internal parameter calculation, and obtain the optimal parameters, which specifically includes: Step S31: Establish a laser radar internal parameter calibration model, where the internal parameter is in the form of a calibration matrix: In the above formula, R represents the 3D rotation matrix, t is the translation vector, s is the scale factor; for any measurement value x, it is converted into another 3D Cartesian coordinate through the H matrix: ； Step 32: construct an internal parameter optimization model, assuming that the sampling point cloud of a specific calibration target plate calibration object under a specific internal parameter configuration is , T represents the number of scanning postures at the same scanning position; the optimization loss is defined as: In the above formula, N is the number of sampling locations, M is the number of calibration target plates for a single sampling, That is, the average error between the distances between the coordinates of the cone tops of different cones on the same calibration target obtained in step S24 and the true value; J Implicitly expresses the error relationship between the distance between the internal parameter matrix H and the characteristic points on the calibration target and the true value, by optimizing J , estimate the optimal value of H; Step S33, based on the heuristic optimization algorithm, through iterative optimization, solve the optimal estimate of the internal parameter matrix; Step S4, performance verification step: based on the calculated optimal parameters, the distance between the markers on the verification target is obtained to verify the performance after calibration.

2. A laser radar internal parameter calibration method according to claim 1, characterized in that: The data acquisition step in step S1 specifically includes: Step S11: configuring the initial internal parameters of the laser radar according to the design parameters; Step S12: Install the laser radar on a movable calibration platform; Step S13: Control the calibration platform to move to each calibration test point in sequence according to the calibration test point sequence; Step S14: At each calibration test point, adjust the angle between the laser radar and the horizontal plane by adjusting the angle bracket, sample the calibration targets at different positions in the calibration environment, and obtain point cloud data under the current parameter configuration; Step S15: Complete data sampling at all calibration test points and store the acquired data.

3. The laser radar internal parameter calibration method according to claim 1, characterized in that: The performance verification step in step S4 specifically includes: Step S41: Based on the optimized internal parameters, the laser radar is configured to scan two calibration target plates that are not within the original calibration range to obtain corresponding sampling point clouds; Step S42, then using the method of step S24, obtain the errors between the coordinates of the cone apexes of each cone; Step S43: If the error is less than the threshold d TH , the calibration process is terminated and the calibration parameters are output; Step S44: If the error is greater than the threshold d TH , then re-collect the calibration data and re-calibrate the parameters.

4. A device for laser radar internal parameter calibration method, characterized in that: A laser radar internal parameter calibration method according to any one of claims 1 to 3 is used, wherein the device comprises: Calibration target, used for calibration in combination with cone markers; Verification targets, used to obtain data at longer distances than the calibration targets; Calibration platform, used to provide a lidar calibration test platform; Communication module, used to transmit lidar data and calibration processing results of the computing system; A computing system, configured to send motion control instructions to the calibration platform through a communication module according to the positions of calibration test points; The storage server is used to receive the lidar data and lidar calibration results sent by the computing system, and store the lidar data and lidar calibration results in the form of files.

5. The device for calibrating the internal parameters of a laser radar according to claim 4, characterized in that: The calibration platform includes a laser radar, an angle adjustment bracket, a mobile platform, and a calibration platform controller.

6. The device for calibrating the internal parameters of a laser radar according to claim 4, characterized in that: The communication module includes a wireless serial port, a wireless network port, a first network port, and a second network port.

Images