GNSS-assisted accurate robust camera-IMU calibration method and system

Abstract

The invention discloses a GNSS-assisted accurate robust camera-IMU calibration method and system, and belongs to the technical field of sensor calibration, and the method comprises the steps: employing a visual mileage calculation method to obtain a camera pose, and employing a GNSS / INS loose combination method to calculate and obtain an IMU pose; constructing a camera-IMU rotation extrinsic parameter relational expression and a camera-IMU translation extrinsic parameter relational expression; based on the camera-IMU rotation extrinsic parameter relational expression, performing preliminary solution by adopting least square and inter-epoch iteration, and then performing preliminary solution by adopting least square and iteration in combination with the camera-IMU translation extrinsic parameter relational expression; and based on the camera-IMU rotation external parameter relational expression and the camera-IMU translation external parameter relational expression after preliminary solution, performing observable external parameter optimization by adopting a data screening method and a truncated singular value decomposition method to obtain exact solutions of the camera-IMU rotation external parameter and the camera-IMU translation external parameter, and completing calibration. According to the method, IMU error drift is effectively inhibited, sections with poor observability are filtered out, and the overall observation quality and the calibration precision and robustness are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of sensor calibration, and in particular relates to an accurate and robust camera-IMU calibration method and system assisted by GNSS. Background Art

[0002] Accurate calibration of cameras and inertial measurement units (IMUs) is not only the basis for the effective fusion of image information and inertial data, but also a key link in the field of multi-sensor fusion, and is widely used in high-precision navigation, attitude estimation, robot perception, unmanned systems, etc. However, current calibration methods still face many challenges in terms of accuracy, stability, and the ability to adapt to complex environments.

[0003] Existing camera-IMU calibration methods are mainly divided into external marker-based calibration and odometer-based calibration. The external marker-based method relies on specific visual markers such as Apriltag QR codes to provide reference points for the camera and IMU, but its dependence on external features and the cumbersome experimental process limit its applicability. The odometer-based calibration method can complete the calibration in a featureless environment, but due to the zero bias error of the IMU and the uncertainty of the monocular visual scale, the calibration accuracy of this type of method is easily affected. In addition, blindly processing all available data may not only use invalid observations such as poor observability and gross errors, but also lead to unnecessary high computational costs. This patent pays special attention to data with poor observability. For example, commonly used mobile platforms (such as unmanned vehicles, robots, unmanned ships, etc.) may experience observation degradation when moving in the vertical direction and around the horizontal axis, resulting in insufficient observability and limited stability and accuracy of the solution.

[0004] Therefore, it is necessary to design an accurate and robust camera-IMU calibration method and system assisted by the Global Navigation Satellite System (GNSS) to improve positioning accuracy in order to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to address the problems existing in the prior art and provide an accurate and robust camera-IMU calibration method and system using GNSS assistance. By introducing GNSS assistance, a data screening method based on information theory and a truncated singular value decomposition (TSVD) method are proposed, which not only improves the accuracy and quality of the data, but also effectively solves the problem of inhomogeneous observation data in complex environments.

[0006] According to one aspect of this specification, there is provided a precise and robust camera-IMU calibration method using GNSS assistance, comprising:

[0007] The visual odometry method is used to obtain the camera pose, and the GNSS / INS loose combination method is used to obtain the IMU pose. The two poses are respectively obtained by inter-epoch difference to obtain their respective inter-epoch relative poses.

[0008] Based on the relative poses between their respective epochs, the camera-IMU rotation extrinsic parameter relationship and the camera-IMU translation extrinsic parameter relationship are constructed;

[0009] Based on the camera-IMU rotation extrinsic parameter relationship, the least square method and epoch iteration are used to preliminarily solve the camera-IMU rotation extrinsic parameter. Combined with the result of the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship, the least square calculation and iteration are used to preliminarily solve the camera-IMU translation extrinsic parameter.

[0010] Based on the camera-IMU rotation extrinsic parameter relationship after the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship after the preliminary solution of the camera-IMU translation extrinsic parameter, the data screening method and the truncated singular value decomposition method are used to optimize the observable extrinsic parameters, and the accurate solutions of the camera-IMU rotation extrinsic parameters and the camera-IMU translation extrinsic parameters are obtained to complete the calibration.

[0011] Furthermore, the least square method and epoch iteration are used to preliminarily solve the camera-IMU rotation external parameters, including:

[0012] The rotation matrix in the camera-IMU rotation external parameter relationship is converted into a quaternion, and the quaternion is solved by least squares;

[0013] A rotation equation group is constructed based on K moments, and the rotation equation group is solved through iteration between epochs to obtain a preliminary solution of the camera-IMU rotation extrinsic parameters.

[0014] Furthermore, the least squares and iteration are used to preliminarily solve the camera-IMU translation external parameters, including:

[0015] The camera-IMU translation extrinsic parameter relationship is converted into a translation equation group, and the calculation formula is obtained by least squares calculation. The calculation formula is iteratively solved based on the preliminary solution of the camera-IMU rotation extrinsic parameter to obtain the preliminary solution of the camera-IMU translation extrinsic parameter.

[0016] Furthermore, data screening method and truncated singular value decomposition method are used to optimize the external parameters of observability, including:

[0017] Based on the camera-IMU rotation extrinsic parameter relationship after the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship after the preliminary solution of the camera-IMU translation extrinsic parameter, the normal equation and the Fisher information matrix are constructed to perform data screening;

[0018] Based on the data segments after data screening, the truncated singular value decomposition method is used to obtain the singular values ​​that meet the threshold limit for solution, and the optimized camera-IMU rotation extrinsic parameters and camera-IMU translation extrinsic parameters are obtained.

[0019] Further, data screening is performed, including:

[0020] Calculate the minimum singular value of the Fisher information matrix of different data segments and filter out the data segment with the largest amount of information.

[0021] According to one aspect of the present specification, there is provided an accurate and robust camera-IMU calibration system using GNSS assistance, comprising:

[0022] The pose acquisition module is used to obtain the camera pose using the visual mileage calculation method, and the IMU pose using the GNSS / INS loose combination method. The two poses are respectively obtained by inter-epoch difference to obtain their respective inter-epoch relative poses;

[0023] The module for constructing extrinsic relations is used to construct camera-IMU rotation extrinsic relations and camera-IMU translation extrinsic relations based on their relative poses between epochs.

[0024] The preliminary solution module is used to perform preliminary solution of the camera-IMU rotation external parameters based on the camera-IMU rotation external parameter relationship, using least squares and epoch iteration, and combining the results of the preliminary solution of the camera-IMU rotation external parameters and the camera-IMU translation external parameter relationship, and using least squares calculation and iteration to perform preliminary solution of the camera-IMU translation external parameters;

[0025] The extrinsic parameter optimization and calibration module is used to optimize the observable extrinsic parameters based on the camera-IMU rotation extrinsic parameter relationship after the preliminary solution of the camera-IMU rotation extrinsic parameters and the camera-IMU translation extrinsic parameter relationship after the preliminary solution of the camera-IMU translation extrinsic parameters, and obtain the accurate solutions of the camera-IMU rotation extrinsic parameters and the camera-IMU translation extrinsic parameters to complete the calibration.

[0026] According to one aspect of the present specification, there is provided an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the precise and robust camera-IMU calibration method using GNSS assistance when executing the computer program.

[0027] According to one aspect of the present specification, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that when the computer program is executed by a processor, the steps of the precise and robust camera-IMU calibration method using GNSS assistance are implemented.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. The present invention proposes a precise and robust camera-IMU calibration method and system using GNSS assistance, which calculates the IMU pose through a GNSS / INS loose combination algorithm, effectively suppresses the error drift of the IMU during long-term use and improves the calibration accuracy.

[0030] 2. The present invention proposes an accurate and robust camera-IMU calibration method and system using GNSS assistance, which adopts a data screening method based on information theory to divide the camera and IMU observation values ​​into different segments, effectively filtering out segments with poor observability and improving the overall observation quality.

[0031] 3. The present invention proposes a precise and robust camera-IMU calibration method and system using GNSS assistance, which uses a truncated singular value decomposition method to screen singular values, effectively reducing the impact of low-quality data on calculation results and improving the accuracy and robustness of calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 This is a flow chart of a camera-IMU calibration method according to an embodiment of the present invention;

[0034] Figure 2 A schematic diagram of camera-IMU external parameters according to an embodiment of the present invention;

[0035] Figure 3 Schematic diagram of the combined posture of VO and GNSS / INS according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0037] The present invention provides an accurate and robust camera-IMU calibration method using GNSS assistance, such as Figure 1 As shown, it includes: using visual mileage calculation method to obtain camera pose, using GNSS / INS loose combination method to obtain IMU pose, and obtaining the relative pose between epochs of the two poses through epoch-to-epoch differences. Based on the relative poses between epochs, the camera-IMU rotation extrinsic parameter relationship and the camera-IMU translation extrinsic parameter relationship are constructed. Based on the camera-IMU rotation extrinsic parameter relationship, the camera-IMU rotation extrinsic parameter is initially solved by least squares and epoch-to-epoch iteration. Combined with the results of the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship, the camera-IMU translation extrinsic parameter is initially solved by least squares and iteration. Based on the camera-IMU rotation extrinsic parameter relationship after the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship after the preliminary solution of the camera-IMU translation extrinsic parameter, the data screening method and the truncated singular value decomposition method are used to optimize the observable extrinsic parameters, and the accurate solutions of the camera-IMU rotation extrinsic parameters and the camera-IMU translation extrinsic parameters are obtained to complete the calibration.

[0038] Specifically, suppose the pose The position of coordinate system a in coordinate system b is ,in Represents the rotation of the three axes of coordinates, represents the translation of the origin of the coordinate system. Assume that the camera-IMU external parameters are , where I represents the IMU coordinate system and C represents the camera coordinate system. Calibration is to determine The value of Figure 2 shown.

[0039] Specifically, the camera pose is calculated using the open source visual odometry (VO) algorithm ORB-SLAM2 and is expressed in the world coordinate system (World, W system) as The pose is calculated using the GNSS / INS loose combination method, that is, the GNSS RTK position result is combined with the IMU and expressed in the navigation coordinate system (Navigation, N system) as The definitions of the W and N systems are different. In addition, due to the existence of camera-IMU external parameters, the camera trajectory and IMU trajectory do not overlap, such as Figure 3Although the camera and IMU trajectories do not overlap, the shapes of their trajectories are similar, but they differ in translation by one scale.

[0040] Specifically, for Tie to The camera pose is and posture The transformation relationship is as follows:

[0041] (1)

[0042] For two consecutive epochs , the pose relationship between the camera and IMU is as follows:

[0043] (2)

[0044] (3)

[0045] Invert both sides of equation (2) to obtain:

[0046] (4)

[0047] Multiply the left and right sides of equation (4) and equation (3) respectively to obtain:

[0048]

[0049] (5)

[0050]

[0051] This eliminates the unknown , and we get The hand-eye calibration relationship of Written as rotation and pan In combined form, we can get:

[0052] (6)

[0053] in, Represents the monocular visual scale. Expanding equation (6), deriving it, and extracting the corresponding relationship between rotation and translation, we can obtain:

[0054] (7)

[0055] (8)

[0056] Equations (7) and (8) are derived from the hand-eye calibration relationship and include camera-IMU external parameters , The relationship between and basis.

[0057] Specifically, according to formula (7), let , , which represent consecutive epochs. The camera and IMU rotate relative to each other. Moving to the left we get:

[0058] (9)

[0059] Formula (9) is also The form of is the hand-eye calibration relation about rotation. It belongs to the special orthogonal group SO(3) on the manifold, which is not easy to solve directly. Written as quaternion In the form of:

[0060] (10)

[0061] in, , represent The identity matrix, Represents quaternion The real part of Represents quaternion The imaginary part of vector, Representation vector Constructed antisymmetric matrix.

[0062] According to formula (10), At this moment, the equations can be constructed as follows:

[0063] (11)

[0064] Among them, quaternions By solving the weighted least squares problem, we get:

[0065] (12)

[0066] Specifically, in the actual solution, each new observation of an epoch, that is, the expansion of the equations And solve it again, we get:

[0067] (13)

[0068] The camera-IMU relative rotation relationship is accurately solved by iteration between epochs. Convert to rotation matrix , which is used for the next step of translation parameters Estimates.

[0069] Specifically, if Figure 3 As shown in Figure 2, since the monocular camera cannot provide accurate scale information, The translation given by The translation given by the combined navigation will have scale differences, where the red, green, and blue colors of the coordinate axis represent the corresponding sensor Axis. , After obtaining two sets of poses, according to the formula 8) The translation parts of the two are related as follows:

[0070] (14)

[0071] in, , formula (14) contains the scale and pan Two unknown parameters are written as follows:

[0072] (15)

[0073] Formula (15) can be simplified as , A sequence of key frames can be composed The system of equations can be calculated using least squares as follows:

[0074] (16)

[0075] In the actual solution, the relative rotation between camera and IMU is solved in each iteration Then, based on the historical observations, a new translation observation equation is added for iterative solution to obtain The exact solution of .

[0076] Specifically, in order to overcome the degradation that may occur when there is not enough motion excitation or scene constraints, and to effectively reduce the computational cost, the method proposed in the embodiment of the present invention automatically selects the data segment with the largest amount of information for solution. Assuming that the measurements are independent of each other and contain only Gaussian white noise, the parameter estimation is modeled as a maximum likelihood estimation (MLE) problem, which can be reformulated as follows:

[0077] (17)

[0078] in, , Indicates that from The residual introduced in the measurement has a covariance of . It is a diagonal matrix, and the elements on the diagonal Representative The weight of the observation equation is calculated as ,in represents the exponential function, is the scaling factor.

[0079] Specifically, the embodiment of the present invention introduces the variance matrix into the observation equation of formula (17): (i.e., weights) and is solved using an iterative method (such as the Gaussian-Newton method or the Levenberg-Marquard method). The initial value of is the least squares estimate from the previous step, and in each iteration Linearize the above nonlinear system at: ,in is the Jacobian matrix. From this, we can get the update increment (or error state estimate) The linear least squares problem is as follows:

[0080] (18)

[0081] The optimal solution is given by the normal equation:

[0082] (19)

[0083] Once calculated , you can The estimate is updated to ,in Represents the addition operation on the manifold.

[0084] Specifically, the embodiment of the present invention proposes data segment screening based on information theory. In formula (19): is the Fisher information matrix of the MLE problem (Eq. 17), which contains all the information in the measurement. In the motion of unmanned vehicles, due to the lack of motion incentives and / or environmental constraints, So, collect information data to ensure Full rank (or observability) is crucial. To do this, we first divide the entire data into multiple segments of constant time length, and then calculate the The matrix performs singular value decomposition and omits the subscripts to obtain:

[0085] (20)

[0086] in, is a diagonal matrix of singular values ​​arranged in descending order, , is an orthogonal matrix. The information measure (or observability index) of a single segment based on singular values ​​is adopted, because a larger information measure means more information segments. The expression for screening is as follows:

[0087] (twenty one)

[0088] in, and are the thresholds for the maximum and minimum singular values, respectively. Therefore, for a particular sequence of collected data, the information metric of each segment is first evaluated, and then the segment with the most information is selected for optimization.

[0089] Specifically, the embodiment of the present invention also proposes an observability-based state update. In order to make the proposed camera-IMU calibration robust to (weak) unobservability, the observability constraint is explicitly enforced at each iteration, that is, only the state that falls in the observable direction is updated. To this end, the embodiment of the present invention adopts TSVD to perform a low-rank approximation of the information matrix, that is, at the information threshold selected as a design choice Select singular values ​​above Based on the observability guarantee information matrix and the SVD-based matrix pseudo-inverse, the solution is as follows:

[0090] (twenty two)

[0091] Therefore, it is the same as solving for all singular values ​​of the full information matrix In contrast (which may incorrectly introduce observable directions due to numerical issues), the solution using Eq. (22) explicitly avoids updating unobservable (or weakly observable) parameters, and thus can conservatively ensure observability.

[0092] The implementation basis of each embodiment of the present invention is implemented by programmed processing through a device with a processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this reality, on the basis of the above embodiments, an embodiment of the present invention provides an accurate and robust camera-IMU calibration system using GNSS assistance, which is used to execute an accurate and robust camera-IMU calibration method using GNSS assistance in the above method embodiment.

[0093] The system includes a posture acquisition module, which is used to obtain the camera posture by using the visual mileage calculation method, and to calculate the IMU posture by using the GNSS / INS loose combination method, and to obtain the relative posture between epochs of the two postures by inter-epoch difference;

[0094] The module for constructing external parameter relations is used to construct the camera-IMU rotation external parameter relations and the camera-IMU translation external parameter relations based on their respective relative poses between epochs. The module for preliminary solution is used to perform preliminary solution of the camera-IMU rotation external parameters based on the camera-IMU rotation external parameter relations by using least squares and iteration between epochs, and preliminary solution of the camera-IMU translation external parameters is performed by using least squares calculation and iteration based on the results of preliminary solution of the camera-IMU rotation external parameters and the camera-IMU translation external parameter relations. The module for external parameter optimization and calibration is used to perform observable external parameter optimization based on the camera-IMU rotation external parameter relations and the camera-IMU translation external parameter relations after preliminary solution of the camera-IMU rotation external parameters by using data screening method and truncated singular value decomposition method, and obtain accurate solutions of the camera-IMU rotation external parameters and the camera-IMU translation external parameters to complete calibration.

[0095] An embodiment of the present invention provides an accurate and robust camera-IMU calibration system using GNSS assistance. It aims at the problems of poor observability of data and low accuracy and stability of calibration methods. The above-mentioned modules are used to divide the camera and IMU observation values ​​into different segments, and the truncated singular value decomposition method is used to screen the singular values. This not only effectively filters out the segments with poor observability, suppresses the error drift of the IMU during long-term use, but also improves the accuracy and robustness of the calibration.

[0096] Based on the same inventive concept as the aforementioned embodiment, an embodiment of the present invention further provides an electronic device, including a memory and a processor, the memory being used to store computer-executable instructions, and the processor being used to execute computer-executable instructions, to implement the precise and robust camera-IMU calibration method using GNSS assistance as proposed in the aforementioned embodiment.

[0097] The embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by the processor, it is used to filter out sections with poor observability, improve the overall observation quality and the accuracy and robustness of the calibration. The storage medium can be any non-volatile storage device such as a hard disk, a solid-state drive, a flash drive, an optical disk, etc., for storing computer program codes and necessary data files. The stored computer program includes: a posture acquisition module, an external parameter relationship construction module, a preliminary solution module, and an external parameter optimization and calibration module.

[0098] Finally, it should be pointed out that the above specific embodiments are only representative examples of the present invention. Obviously, the present invention is not limited to the above specific embodiments, and there are many variations. Any simple modification, equivalent changes and modifications made to the above specific embodiments based on the technical essence of the present invention should be considered to belong to the protection scope of the present invention.

Claims

1. An accurate and robust camera-IMU calibration method using GNSS assistance, characterized in that: include: The visual odometry method is used to obtain the camera pose, and the GNSS / INS loose combination method is used to obtain the IMU pose. The two poses are respectively obtained by inter-epoch difference to obtain their respective inter-epoch relative poses. Based on the relative poses between their respective epochs, the camera-IMU rotation extrinsic parameter relationship and the camera-IMU translation extrinsic parameter relationship are constructed; Based on the camera-IMU rotation extrinsic parameter relationship, the least square method and epoch iteration are used to preliminarily solve the camera-IMU rotation extrinsic parameter. Combined with the result of the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship, the least square method and iteration are used to preliminarily solve the camera-IMU translation extrinsic parameter. Based on the camera-IMU rotation extrinsic parameter relationship after the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship after the preliminary solution of the camera-IMU translation extrinsic parameter, the data screening method and the truncated singular value decomposition method are used to optimize the observable extrinsic parameters, and the accurate solutions of the camera-IMU rotation extrinsic parameters and the camera-IMU translation extrinsic parameters are obtained to complete the calibration.

2. The method of accurate and robust camera-IMU calibration using GNSS assistance according to claim 1, characterized in that: The camera-IMU rotation external parameters are initially solved using least squares and epoch iteration, including: The rotation matrix in the camera-IMU rotation external parameter relationship is converted into a quaternion, and the quaternion is solved by least squares; A rotation equation group is constructed based on K moments, and the rotation equation group is solved through iteration between epochs to obtain a preliminary solution of the camera-IMU rotation extrinsic parameters.

3. The method of accurate and robust camera-IMU calibration using GNSS assistance according to claim 2, characterized in that: The camera-IMU translation extrinsic parameters are initially solved using least squares and iteration, including: The camera-IMU translation extrinsic parameter relationship is converted into a translation equation group, and the calculation formula is obtained by least squares calculation. The calculation formula is iteratively solved based on the preliminary solution of the camera-IMU rotation extrinsic parameter to obtain the preliminary solution of the camera-IMU translation extrinsic parameter.

4. The method of accurate and robust camera-IMU calibration using GNSS assistance according to claim 1, characterized in that: Data screening method and truncated singular value decomposition method are used to optimize external parameters of observability, including: Based on the camera-IMU rotation extrinsic parameter relationship after the preliminary solution of the camera-IMU rotation extrinsic parameter and the camera-IMU translation extrinsic parameter relationship after the preliminary solution of the camera-IMU translation extrinsic parameter, the normal equation and the Fisher information matrix are constructed to perform data screening; Based on the data segments after data screening, the truncated singular value decomposition method is used to obtain the singular values ​​that meet the threshold limit for solution, and the optimized camera-IMU rotation extrinsic parameters and camera-IMU translation extrinsic parameters are obtained.

5. The method of accurate and robust camera-IMU calibration using GNSS assistance according to claim 4, characterized in that: Perform data screening, including: Calculate the minimum singular value of the Fisher information matrix of different data segments and filter out the data segment with the largest amount of information.

6. An accurate and robust camera-IMU calibration system using GNSS assistance, characterized in that: include: The pose acquisition module is used to obtain the camera pose using the visual mileage calculation method, and the IMU pose using the GNSS / INS loose combination method. The two poses are respectively obtained by inter-epoch difference to obtain their respective inter-epoch relative poses; The module for constructing extrinsic relations is used to construct camera-IMU rotation extrinsic relations and camera-IMU translation extrinsic relations based on their relative poses between epochs. The preliminary solution module is used to perform preliminary solution of the camera-IMU rotation external parameters based on the camera-IMU rotation external parameter relationship, using least squares and epoch iteration, and combining the results of the preliminary solution of the camera-IMU rotation external parameters and the camera-IMU translation external parameter relationship, and using least squares calculation and iteration to perform preliminary solution of the camera-IMU translation external parameters; The extrinsic parameter optimization and calibration module is used to optimize the observable extrinsic parameters based on the camera-IMU rotation extrinsic parameter relationship after the preliminary solution of the camera-IMU rotation extrinsic parameters and the camera-IMU translation extrinsic parameter relationship after the preliminary solution of the camera-IMU translation extrinsic parameters, and obtain the accurate solutions of the camera-IMU rotation extrinsic parameters and the camera-IMU translation extrinsic parameters to complete the calibration.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the accurate and robust camera-IMU calibration method using GNSS assistance as described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the accurate and robust camera-IMU calibration method using GNSS assistance as described in any one of claims 1 to 5 are implemented.

Images