Automatic calibration method, system and device for laser radar parameters

Abstract

This application relates to the field of lidar system calibration technology, specifically to a method, system, and apparatus for automatic parameter calibration of lidar. The method includes: acquiring rotational angular velocity, steering angle data, and point cloud data of an autonomous vehicle; establishing a vehicle coordinate system and a lidar coordinate system; constructing the lidar's extrinsic parameter matrix, longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence; determining the motion constraint strength and calibration constraint reliability at the acquisition time; and updating the extrinsic parameter matrix at the acquisition time based on the calibration constraint reliability and the extrinsic parameter matrix, thereby achieving automatic parameter calibration of the lidar. This application can improve the accuracy of automatic parameter calibration of lidar.

Description

Technical Field

[0001] This application relates to the field of calibration technology for lidar systems, and specifically to automatic parameter calibration methods, systems and devices for lidar. Background Technology

[0002] LiDAR is primarily used for environmental perception in autonomous driving systems. During LiDAR operation, automatic parameter calibration is a crucial foundation for ensuring its accuracy. The raw point cloud data acquired by LiDAR must be calibrated to be transformed into spatially meaningful vehicle coordinate system data. Incorrect calibration can lead to deviations in obstacle localization during autonomous driving, causing serious safety hazards. During automatic parameter calibration, the LiDAR's extrinsic parameter matrix uses rigid transformation to convert the obstacle's point cloud coordinates from the radar coordinate system to the vehicle coordinate system. Current technology mainly relies on the diverse longitudinal and lateral movements of the vehicle to provide strong constraints on the extrinsic parameter matrix, enabling the vehicle to accurately locate obstacles.

[0003] However, the relative motion information collected in low-speed or single-motion modes of the vehicle is insufficient, making it difficult to generate enough relative pose change information. This makes it impossible to provide strong constraints for the extrinsic parameter matrix and to implement precise rigid transformations to convert obstacles into the vehicle coordinate system. As a result, the position of the converted obstacle point cloud in the vehicle coordinate system deviates from its actual position, which can easily lead to distortion of the vehicle's environmental perception and reduce the safety and reliability of autonomous driving. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide an automatic parameter calibration method, system, and apparatus for lidar, the specific technical solution of which is as follows:

[0005] In a first aspect, embodiments of this application provide an automatic parameter calibration method for lidar, the method comprising the following steps:

[0006] The system collects rotational angular velocity, steering angle data, and point cloud data of autonomous vehicles at different acquisition times, establishes a vehicle coordinate system and a lidar coordinate system, and constructs the lidar extrinsic parameter matrix for each acquisition time using the point cloud data, lidar coordinate system, and vehicle coordinate system. Based on the rotational angular velocity and steering angle data, the system constructs the longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence for each acquisition time.

[0007] Based on the principal component information of the longitudinal and lateral displacement sequences at the same acquisition time, and the degree of nonlinear correlation between the angular velocity sequence and the lateral and longitudinal displacement sequences respectively, the motion constraint strength at the same acquisition time is determined. Based on the difference in motion constraint strength at the acquisition time and all acquisition times before the acquisition time and the probability distribution weighted by exponential decay, the calibration constraint reliability at the acquisition time is determined.

[0008] Based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix of the acquisition time adjacent to the acquisition time, the extrinsic parameter matrix at the acquisition time is updated to realize the automatic calibration of the lidar parameters.

[0009] Furthermore, the process for determining the longitudinal displacement sequence and the transverse displacement sequence is as follows:

[0010] The integral of the rotational angular velocity at two adjacent data acquisition moments is used as the total displacement of the wheel motion during the time interval corresponding to the two adjacent data acquisition moments; based on the steering angle data and the total displacement of the wheel motion, the longitudinal displacement component and the lateral displacement component of the time interval corresponding to the two adjacent data acquisition moments are calculated.

[0011] Based on the longitudinal displacement components at each time interval, a longitudinal displacement sequence at the acquisition time is established.

[0012] Based on the lateral displacement components at each time interval, a lateral displacement sequence at the acquisition time is established.

[0013] Furthermore, the process of constructing the motion constraint strength at the same acquisition moment is as follows:

[0014] Obtain the eigenvalues ​​of the first principal components of the longitudinal and lateral displacement sequences at the same acquisition time;

[0015] Calculate the mutual information between the angular velocity sequence at the same acquisition time and the lateral displacement sequence and the longitudinal displacement sequence, respectively, and calculate the first mean value at the same acquisition time based on the mutual information.

[0016] The positive correlation between the first mean and the eigenvalue of the first principal component at the same acquisition time is denoted as the motion constraint intensity at the same acquisition time.

[0017] Furthermore, the process of obtaining the first mean is as follows:

[0018] The mean of the mutual information between the angular velocity sequence and the lateral displacement sequence and the longitudinal displacement sequence at the same acquisition time is denoted as the first mean at the same acquisition time.

[0019] Furthermore, the process for determining the reliability of the calibration constraints at the acquisition time is as follows:

[0020] The second mean value of the acquisition time is determined based on the motion constraint intensity of all acquisition times within the first preset time period before the acquisition time.

[0021] The probability distribution of the constraint strength at the acquisition time, weighted by exponential decay, is denoted as the effective motion weight at the acquisition time.

[0022] The difference between the number 1 and the tangent of the second mean at the acquisition time is recorded as the first difference at the acquisition time. The positive correlation between the first difference at the acquisition time and the effective motion weight is recorded as the calibration constraint reliability at the acquisition time.

[0023] Furthermore, the process for determining the second mean is as follows:

[0024] The coefficient of variation of the motion constraint intensity at all acquisition times within the first preset time period before the acquisition time is denoted as the motion constraint coefficient of variation at the acquisition time, and the mean of the motion constraint coefficients of variation at all acquisition times before the acquisition time is denoted as the second mean at the acquisition time.

[0025] Furthermore, the specific process of updating the extrinsic parameter matrix at the acquisition time based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix at acquisition times adjacent to the acquisition time, is as follows:

[0026] Based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix of the acquisition time adjacent to the acquisition time, the extrinsic parameter fusion matrix at the acquisition time is determined.

[0027] The extrinsic parameter fusion matrix at the acquisition time is used as the updated matrix of the extrinsic parameter matrix at the acquisition time.

[0028] Furthermore, the process of determining the extrinsic parameter fusion matrix is ​​as follows:

[0029] The reliability of the calibration constraint at the acquisition time is used as the weight of the extrinsic parameter matrix at the acquisition time. The difference between the number 1 and the reliability of the calibration constraint at the acquisition time is used as the weight of the extrinsic parameter matrix of the previous acquisition time. The extrinsic parameter matrices of the acquisition time and the previous acquisition time are weighted and summed, and denoted as the extrinsic parameter fusion matrix at the acquisition time.

[0030] Secondly, embodiments of this application provide an automatic parameter calibration device for lidar, the automatic parameter calibration device comprising: a calibration data acquisition module, a constraint reliability evaluation module, and an automatic parameter calibration module.

[0031] The calibration data acquisition module is used to collect the rotational angular velocity, steering angle data, and point cloud data of the autonomous vehicle at different acquisition times, establish the vehicle coordinate system and the lidar coordinate system, and construct the lidar extrinsic parameter matrix at each acquisition time using the point cloud data, lidar coordinate system, and vehicle coordinate system. Based on the rotational angular velocity and steering angle data, the longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence at each acquisition time are constructed respectively.

[0032] The constraint reliability evaluation module is used to determine the motion constraint strength at the same acquisition time based on the principal component information of the longitudinal and lateral displacement sequences at the same acquisition time, as well as the degree of nonlinear correlation between the angular velocity sequence and the lateral and longitudinal displacement sequences, respectively. Based on the difference in motion constraint strength at the acquisition time and all acquisition times before the acquisition time and the probability distribution weighted by exponential decay, the module determines the calibration constraint reliability at the acquisition time.

[0033] The automatic parameter calibration module is used to update the extrinsic parameter matrix at the acquisition time based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix at the acquisition time adjacent to the acquisition time, so as to realize the automatic parameter calibration of the lidar.

[0034] Thirdly, embodiments of this application also provide an automatic parameter calibration system for lidar, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0035] As can be seen from the above embodiments, the automatic parameter calibration method, system, and apparatus for lidar provided in this application have at least the following beneficial effects:

[0036] This application first constructs the extrinsic parameter matrix of the LiDAR at each acquisition time. The extrinsic parameter matrix as a whole describes the spatial relationship between the LiDAR coordinate system and the vehicle coordinate system. Then, due to insufficient relative motion information of the vehicle in low-speed or single-motion modes, it is difficult to generate sufficient relative pose change information, often resulting in insufficient constraints required by the extrinsic parameter matrix. Therefore, the richness of the vehicle's pose change at the acquisition time is analyzed to obtain the strength of the kinematic constraints provided by the information acquired by the LiDAR at the acquisition time for the LiDAR extrinsic parameter calibration, and to obtain the motion constraint strength at the acquisition time. Furthermore, considering that the parameter jitter of the sensor caused by mechanical vibration in low-speed driving or single-motion modes will further lead to inaccurate calculation results, the calibration... The reliability of calibration constraints is difficult to estimate and is prone to errors in extrinsic parameters. Therefore, the reliability of the effective constraints provided to the extrinsic parameter matrix at each acquisition time is quantified to obtain the calibration constraint reliability at each acquisition time. Finally, based on the calibration constraint reliability and extrinsic parameter matrix at each acquisition time, as well as the extrinsic parameter matrices of adjacent acquisition times, the extrinsic parameter matrix at each acquisition time is updated. This achieves automatic parameter calibration of the LiDAR, addressing the problem of insufficient relative motion information acquired by LiDAR in low-speed or single-motion modes of vehicles, resulting in insufficient constraints on the extrinsic parameter matrix, distorted environmental perception, and reduced safety and reliability of autonomous driving. This improves the accuracy of automatic parameter calibration of LiDAR, thereby enhancing the safety and reliability of autonomous driving. Attached Figure Description

[0037] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating the steps of an automatic parameter calibration method for lidar provided in one embodiment of this application;

[0039] Figure 2 This is a schematic diagram of an automatic parameter calibration device for lidar provided in one embodiment of this application. Detailed Implementation

[0040] To further illustrate the technical means and effects adopted by this application in order to achieve the intended purpose of the invention, the following detailed description, in conjunction with the accompanying drawings and preferred embodiments, describes the specific implementation, structure, features and effects of the automatic parameter calibration method, system and device for lidar proposed in this application.

[0041] The following description, in conjunction with the accompanying drawings, details the specific scheme of the automatic parameter calibration method, system, and device for lidar provided in this application.

[0042] Please see Figure 1 The diagram illustrates a flowchart of an automatic parameter calibration method for lidar provided in one embodiment of this application. The method includes the following steps:

[0043] S001: Collect rotational angular velocity, steering angle data, and point cloud data of autonomous vehicles at different collection times, establish vehicle coordinate system and LiDAR coordinate system, construct the LiDAR extrinsic parameter matrix at each collection time using point cloud data, LiDAR coordinate system, and vehicle coordinate system, and construct longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence at each collection time based on rotational angular velocity and steering angle data.

[0044] Wheel speed sensors are installed at the non-steering wheel hubs of autonomous vehicles, directly contacting the tire rotation plane, to collect the wheel's rotational angular velocity; a steering angle sensor is installed at the vehicle's steering column to collect the vehicle's steering angle data; and a LiDAR is installed at the vehicle's center of gravity to scan objects around the vehicle and generate point cloud data, which is used to solve the initial extrinsic parameter matrix.

[0045] Preferably, in one embodiment of this application, when collecting the wheel's rotational angular velocity, steering angular velocity, and point cloud data, the data sampling frequency is 10Hz. In practical applications, as other implementations, the implementer can determine the sampling frequency according to the actual situation, and this application does not impose any special restrictions.

[0046] A vehicle coordinate system is established with the center of the non-steering wheel hub as the origin, and a lidar coordinate system is established with the lidar installation position as the origin.

[0047] By matching the LiDAR coordinate system and the vehicle coordinate system with point cloud data, the initial rotation matrix and translation vector are obtained, and the extrinsic parameter matrix of the LiDAR is constructed. , ,in, Represents the initial rotation matrix; This represents the translation vector.

[0048] The initial rotation matrix and translation vector are obtained by matching the coordinates of corresponding points in the vehicle coordinate system and the LiDAR coordinate system using point cloud data. The initial rotation matrix is ​​a 3×3 orthogonal matrix representing the rotation components; its elements are the direction cosines of the three coordinate axes of the LiDAR coordinate system in the vehicle coordinate system. The translation vector is a 3×1 vector representing the translation components, indicating the three-dimensional position of the LiDAR coordinate system origin in the vehicle coordinate system, i.e., the longitudinal, lateral, and height offsets relative to the rear axle center of the vehicle. The extrinsic parameter matrix... The spatial relationship between the lidar coordinate system and the vehicle coordinate system is described in general. The construction of the initial rotation matrix, translation vector, and extrinsic parameter matrix is ​​a well-known technique and will not be elaborated further.

[0049] By integrating the rotational angular velocity at two adjacent data acquisition moments over time, the total displacement of the wheel motion during the time interval corresponding to the two adjacent data acquisition moments is obtained.

[0050] Based on the Ackermann steering geometry model, the longitudinal and lateral displacement components of the time interval corresponding to two adjacent data acquisition moments are calculated according to the steering angle data and the total displacement of the wheel movement.

[0051] The longitudinal displacement components of the wheel motion at each time interval corresponding to the adjacent data acquisition time before the acquisition time are arranged in chronological order to obtain the longitudinal displacement sequence at the acquisition time; the lateral displacement components of the wheel motion at each time interval corresponding to the adjacent data acquisition time before the acquisition time are arranged in chronological order to obtain the lateral displacement sequence at the acquisition time; and the rotational angular velocities at each data acquisition time before the acquisition time are arranged in chronological order to obtain the angular velocity sequence at the acquisition time.

[0052] Thus, the longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence at each acquisition moment are obtained.

[0053] S002: Based on the principal component information of the longitudinal and lateral displacement sequences at the same acquisition time, and the degree of nonlinear correlation between the angular velocity sequence and the lateral and longitudinal displacement sequences respectively, the motion constraint strength at the same acquisition time is determined. Based on the difference in motion constraint strength at the acquisition time and all acquisition times before the acquisition time, and the probability distribution weighted by exponential decay, the calibration constraint reliability at the acquisition time is determined.

[0054] The vehicle lacks sufficient relative motion information in low-speed or single motion modes, making it difficult to generate enough relative pose change information. This often results in insufficient constraints required by the extrinsic parameter matrix, making it impossible to implement precise rigid transformations to convert obstacles into the vehicle coordinate system.

[0055] Based on the principal component information of the longitudinal and lateral displacement sequences at the same acquisition time, and the degree of nonlinear correlation between the angular velocity sequence and the lateral and longitudinal displacement sequences respectively, the motion constraint strength at the same acquisition time is determined.

[0056] Principal Component Analysis (PCA) is used to process the longitudinal and lateral displacement sequences at the same acquisition time to obtain the eigenvalues ​​of the first principal component at the same acquisition time. Mutual information analysis is used to calculate the mutual information between the angular velocity sequence at the same acquisition time and the lateral and longitudinal displacement sequences, respectively. The mean of the mutual information between the angular velocity sequence at the same acquisition time and the lateral and longitudinal displacement sequences, respectively, is denoted as the first mean at the same acquisition time. The positive correlation result between the first mean at the same acquisition time and the eigenvalues ​​of the first principal component is denoted as the motion constraint strength at the same acquisition time.

[0057] It is understood that a positive correlation is applied to the eigenvalues ​​of the first mean and the first principal component, ensuring that the first mean and the eigenvalues ​​of the first principal component are positively correlated with the motion constraint intensity. It is understood that the positive correlation in this application refers to the relationship between the independent variable and the dependent variable, where the independent variables are the first mean and the eigenvalues ​​of the first principal component, and the dependent variable is the motion constraint intensity. The positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive relationship, a multiplicative relationship, etc.

[0058] Preferably, as an embodiment of this application, the product of the first mean and the eigenvalue of the first principal component at the same acquisition time is denoted as the motion constraint intensity at the same acquisition time.

[0059] Using the PCA principal component analysis algorithm to obtain the eigenvalues ​​of the first principal component and using mutual information analysis to calculate the mutual information degree are both well-known techniques and will not be elaborated further.

[0060] The eigenvalues ​​of the first principal component quantify the intensity of the vehicle's motion trajectory in its main extension directions, reflecting the significance of the vehicle's rotation. A larger eigenvalue indicates richer rotational motion, providing stronger kinematic constraints for lidar extrinsic parameter calibration. The first mean measures the overall correlation between angular velocity and lateral and longitudinal displacements. A larger first mean indicates richer pose changes during vehicle motion, providing stronger kinematic constraints for lidar extrinsic parameter calibration. Therefore, a stronger kinematic constraint indicates richer pose changes at the acquisition time, and stronger kinematic constraints provided by the lidar for lidar extrinsic parameter calibration.

[0061] When a vehicle is traveling at low speed or in a single motion mode, the mechanical vibrations that cause sensor parameter fluctuations can further lead to inaccurate calculation results, making it difficult to estimate the reliability of the calibration and easily generating erroneous external parameters.

[0062] The reliability of the calibration constraints at the acquisition time is determined based on the differences in motion constraint strength at the acquisition time and all acquisition times prior to the acquisition time, and the probability distribution weighted by exponential decay.

[0063] The coefficient of variation of motion constraint intensity at all acquisition times within the first preset time period prior to the acquisition time is denoted as the motion constraint coefficient of variation at the acquisition time. The mean of the motion constraint coefficients of variation at the acquisition time and all acquisition times prior to the acquisition time is denoted as the second mean at the acquisition time. An exponentially decaying weighted probability analysis algorithm is used to process the motion constraint intensity at the acquisition time and all acquisition times prior to the acquisition time, obtaining the probability distribution of the constraint intensity at each acquisition time after exponential decay weighting. This probability distribution is denoted as the effective motion weight at the corresponding acquisition time. The difference between the number 1 and the tangent of the second mean at the acquisition time is denoted as the first difference at the acquisition time. The positive correlation result between the first difference at the acquisition time and the effective motion weight is denoted as the calibration constraint reliability at the acquisition time.

[0064] It is understood that a positive correlation is applied to the first difference at the acquisition time and the effective motion weights, ensuring that both the first difference at the acquisition time and the effective motion weights are positively correlated with the reliability of the calibration constraints. It is also understood that the positive correlation in this application refers to the relationship between the independent and dependent variables, where the independent variables are the first difference at the acquisition time and the effective motion weights, and the dependent variable is the reliability of the calibration constraints. The positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive or multiplicative relationship.

[0065] Preferably, as an embodiment of this application, the product of the first difference at the acquisition time and the effective motion weight is denoted as the calibration constraint reliability at the acquisition time.

[0066] The first preset duration is a preset time length. In this embodiment, the first preset duration is set to 10 seconds. When the time before the acquisition time is less than the first preset duration, the mean-filling method is used to fill in the motion constraint intensity of each acquisition time within the missing time period. The calculation of the coefficient of variation is a well-known technique and will not be elaborated further. The second mean represents the overall relative fluctuation of the motion constraint intensity at the acquisition time and all acquisition times before the acquisition time. The larger the second mean, the more drastic the overall relative fluctuation of the motion constraint intensity, and the more likely there is a sudden braking or turning situation at the acquisition time. The reliability of the motion constraint is low, and the calibration results are easily interfered with. The probability distribution of the data obtained by using the exponential decay weighted probability analysis algorithm is a well-known technique and will not be elaborated further. The effective motion weight represents the weighted proportion of the motion constraint intensity in the recent historical moments of the corresponding acquisition time. The closer the effective motion weight is to 1, the more continuously effective the motion pattern in the recent historical moments of the acquisition time is, and the higher the quality of constraint provided by the acquisition time. Therefore, the reliability of calibration constraints quantifies the reliability of the effective constraints that can be provided to the extrinsic parameter matrix at the corresponding acquisition time. The greater the reliability of calibration constraints at the acquisition time and the closer it is to 1, the more diverse the continuous motion of the vehicle at the acquisition time, and the more sufficient the constraint information can be provided to the extrinsic parameter matrix.

[0067] At this point, the calibration constraint reliability at each acquisition moment is obtained.

[0068] S003: Based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix of the acquisition time adjacent to the acquisition time, the extrinsic parameter matrix at the acquisition time is updated to realize the automatic calibration of the lidar parameters.

[0069] To prevent the lidar from generating incorrect vehicle trajectory due to reliance on incorrect extrinsic parameter matrices, the extrinsic parameter fusion matrix for each acquisition time is determined based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix of the acquisition time adjacent to the acquisition time.

[0070] The reliability of the calibration constraint at the acquisition time is used as the weight of the extrinsic parameter matrix at the acquisition time. The difference between the number 1 and the reliability of the calibration constraint at the acquisition time is used as the weight of the extrinsic parameter matrix of the previous acquisition time. The weighted sum of the extrinsic parameter matrices at the acquisition time and the previous acquisition time is denoted as the extrinsic parameter fusion matrix at the acquisition time.

[0071] The extrinsic parameter fusion matrix at the acquisition time is used as the updated extrinsic parameter matrix at the acquisition time to update the extrinsic parameter matrix at the acquisition time. Rigid body transformation is then performed on the updated extrinsic parameter matrix at the acquisition time to transform the point cloud data acquired by the LiDAR at the acquisition time from the LiDAR coordinate system to the vehicle coordinate system. This transformation of the LiDAR point cloud data from the LiDAR coordinate system to the vehicle coordinate system enables automatic calibration of LiDAR parameters, providing the vehicle with more accurate obstacle detection and environmental perception precision.

[0072] Using the extrinsic parameter fusion matrix at the acquisition time as the updated extrinsic parameter matrix at the acquisition time allows for greater reliance on the current stable extrinsic parameter matrix when the reliability of constraints provided by vehicle motion is higher, and greater reliance on the historical extrinsic parameter matrix when the reliability of constraints provided by vehicle motion is lower. This balances the real-time calibration results with the historical calibration results, avoids extrinsic parameter jumps caused by insufficient vehicle motion information, and enables the extrinsic parameter matrix to be proactively updated when vehicle motion information is sufficient, while relying more on the historical extrinsic parameter matrix for updates when vehicle motion information is invalid.

[0073] This completes the automatic calibration of the lidar parameters.

[0074] Please see Figure 2 , Figure 2 This is a schematic diagram of an automatic parameter calibration device for lidar provided in one embodiment of this application. In this embodiment, the devices include units that perform the steps in the corresponding embodiment of the automatic parameter calibration method for lidar. See also... Figure 2 The automatic parameter calibration device includes: a calibration data acquisition module, a constraint reliability evaluation module, and an automatic parameter calibration module.

[0075] The calibration data acquisition module is used to collect the rotational angular velocity, steering angle data, and point cloud data of the autonomous vehicle at different acquisition times, establish the vehicle coordinate system and the lidar coordinate system, and construct the lidar extrinsic parameter matrix at each acquisition time using the point cloud data, lidar coordinate system, and vehicle coordinate system. Based on the rotational angular velocity and steering angle data, the longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence at each acquisition time are constructed respectively.

[0076] The constraint reliability evaluation module is used to determine the motion constraint strength at the same acquisition time based on the principal component information of the longitudinal and lateral displacement sequences at the same acquisition time, as well as the degree of nonlinear correlation between the angular velocity sequence and the lateral and longitudinal displacement sequences, respectively. Based on the difference in motion constraint strength at the acquisition time and all acquisition times before the acquisition time and the probability distribution weighted by exponential decay, the module determines the calibration constraint reliability at the acquisition time.

[0077] The automatic parameter calibration module is used to update the extrinsic parameter matrix at the acquisition time based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix at the acquisition time adjacent to the acquisition time, so as to realize the automatic parameter calibration of the lidar.

[0078] Based on the same inventive concept as the above methods, this application also provides an automatic parameter calibration system for lidar, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described automatic parameter calibration methods for lidar.

[0079] It is understood that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0080] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0081] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the protection scope of this application.

Claims

1. An automatic parameter calibration method for lidar, characterized in that, The method includes the following steps: The system collects rotational angular velocity, steering angle data, and point cloud data of autonomous vehicles at different acquisition times, establishes a vehicle coordinate system and a lidar coordinate system, and constructs the lidar extrinsic parameter matrix for each acquisition time using the point cloud data, lidar coordinate system, and vehicle coordinate system. Based on the rotational angular velocity and steering angle data, the system constructs the longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence for each acquisition time. Based on the principal component information of the longitudinal and lateral displacement sequences at the same acquisition time, and the degree of nonlinear correlation between the angular velocity sequence and the lateral and longitudinal displacement sequences respectively, the motion constraint strength at the same acquisition time is determined. Based on the difference in motion constraint strength at the acquisition time and all acquisition times before the acquisition time and the probability distribution weighted by exponential decay, the calibration constraint reliability at the acquisition time is determined. Based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix of the acquisition time adjacent to the acquisition time, the extrinsic parameter matrix at the acquisition time is updated to realize the automatic calibration of the lidar parameters. The process of constructing the motion constraint strength at the same acquisition moment is as follows: Obtain the eigenvalues ​​of the first principal components of the longitudinal and lateral displacement sequences at the same acquisition time; Calculate the mutual information between the angular velocity sequence at the same acquisition time and the lateral displacement sequence and the longitudinal displacement sequence, respectively, and calculate the first mean value at the same acquisition time based on the mutual information. The positive correlation between the first mean and the eigenvalue of the first principal component at the same acquisition time is denoted as the motion constraint intensity at the same acquisition time. The process for determining the reliability of the calibration constraints at the acquisition time is as follows: The second mean value of the acquisition time is determined based on the motion constraint intensity of all acquisition times within the first preset time period before the acquisition time. The probability distribution of the constraint strength at the acquisition time, weighted by exponential decay, is denoted as the effective motion weight at the acquisition time. The difference between the number 1 and the tangent of the second mean at the acquisition time is recorded as the first difference at the acquisition time. The positive correlation between the first difference at the acquisition time and the effective motion weight is recorded as the calibration constraint reliability at the acquisition time.

2. The automatic parameter calibration method for lidar as described in claim 1, characterized in that, The process for determining the longitudinal displacement sequence and the transverse displacement sequence is as follows: The integral of the rotational angular velocity at two adjacent data acquisition moments is used as the total displacement of the wheel motion during the time interval corresponding to the two adjacent data acquisition moments; based on the steering angle data and the total displacement of the wheel motion, the longitudinal displacement component and the lateral displacement component of the time interval corresponding to the two adjacent data acquisition moments are calculated. Based on the longitudinal displacement components at each time interval, a longitudinal displacement sequence at the acquisition time is established. Based on the lateral displacement components at each time interval, a lateral displacement sequence at the acquisition time is established.

3. The automatic parameter calibration method for lidar as described in claim 1, characterized in that, The process of obtaining the first mean is as follows: The mean of the mutual information between the angular velocity sequence and the lateral displacement sequence and the longitudinal displacement sequence at the same acquisition time is denoted as the first mean at the same acquisition time.

4. The automatic parameter calibration method for lidar as described in claim 1, characterized in that, The process for determining the second mean is as follows: The coefficient of variation of the motion constraint intensity at all acquisition times within the first preset time period before the acquisition time is denoted as the motion constraint coefficient of variation at the acquisition time, and the mean of the motion constraint coefficients of variation at all acquisition times before the acquisition time is denoted as the second mean at the acquisition time.

5. The automatic parameter calibration method for lidar as described in claim 1, characterized in that, The specific process of updating the extrinsic parameter matrix at the acquisition time based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix at acquisition times adjacent to the acquisition time, is as follows: Based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix of the acquisition time adjacent to the acquisition time, the extrinsic parameter fusion matrix at the acquisition time is determined. The extrinsic parameter fusion matrix at the acquisition time is used as the updated matrix of the extrinsic parameter matrix at the acquisition time.

6. The automatic parameter calibration method for lidar as described in claim 5, characterized in that, The process for determining the extrinsic parameter fusion matrix is ​​as follows: The reliability of the calibration constraint at the acquisition time is used as the weight of the extrinsic parameter matrix at the acquisition time. The difference between the number 1 and the reliability of the calibration constraint at the acquisition time is used as the weight of the extrinsic parameter matrix of the previous acquisition time. The extrinsic parameter matrices of the acquisition time and the previous acquisition time are weighted and summed, and denoted as the extrinsic parameter fusion matrix at the acquisition time.

7. An automatic parameter calibration device for lidar, implementing the method as described in claim 1, characterized in that, The automatic parameter calibration device includes: The calibration data acquisition module is used to collect the rotational angular velocity, steering angle data, and point cloud data of the autonomous vehicle at different acquisition times, establish the vehicle coordinate system and the lidar coordinate system, and construct the lidar extrinsic parameter matrix at each acquisition time using the point cloud data, lidar coordinate system, and vehicle coordinate system. Based on the rotational angular velocity and steering angle data, the longitudinal displacement sequence, lateral displacement sequence, and angular velocity sequence at each acquisition time are constructed respectively. The constraint reliability evaluation module is used to determine the motion constraint strength at the same acquisition time based on the principal component information of the longitudinal and lateral displacement sequences at the same acquisition time, as well as the degree of nonlinear correlation between the angular velocity sequence and the lateral and longitudinal displacement sequences, respectively. Based on the difference in motion constraint strength at the acquisition time and all acquisition times before the acquisition time and the probability distribution weighted by exponential decay, the module determines the calibration constraint reliability at the acquisition time. The automatic parameter calibration module is used to update the extrinsic parameter matrix at the acquisition time based on the calibration constraint reliability and extrinsic parameter matrix at the acquisition time, as well as the extrinsic parameter matrix at the acquisition time adjacent to the acquisition time, so as to realize the automatic parameter calibration of the lidar.

8. An automatic parameter calibration system for lidar, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-6.

Images