A high-precision millimeter-wave radar-camera calibration method and system based on angular reflection

Abstract

The present invention discloses a high-precision millimeter-wave radar-camera calibration method and system based on angular inversion, belonging to the field of sensor calibration technology. The method comprises: mounting a millimeter-wave radar and camera on a mobile carrier, collecting point cloud data and image data from the angular inversion; calculating the distance from the angular inversion to the millimeter-wave radar to obtain the angular inversion point cloud coordinates; performing a grid traversal starting from the angular inversion center point based on the angular inversion image data to obtain the preselected pixel coordinates of the millimeter-wave radar reflection point; obtaining a preliminary estimate of the millimeter-wave radar-camera extrinsic parameters based on the angular inversion point cloud coordinates and pixel coordinates; and iteratively optimizing the extrinsic parameters using a nonlinear optimization algorithm to obtain the final millimeter-wave radar-camera extrinsic parameters, thereby completing the calibration. The present invention effectively increases the number of observations, improves calibration accuracy, reduces the influence of background objects, multipath signals, and other noise, and solves the problem of the difficulty in accurately locating the pixel locations of millimeter-wave signal reflection points from an image.

Description

Technical Field

[0001] The present invention belongs to the technical field of sensor calibration, and in particular relates to a high-precision millimeter-wave radar-camera calibration method and system based on angular reflection. Background Art

[0002] Extrinsic calibration of millimeter-wave radar-camera parameters is a prerequisite for high-precision combined target detection using millimeter-wave radars and cameras. Several millimeter-wave radar-camera calibration methods have been developed based on targets and corner reflectors (corner reflectors). Millimeter-wave signals reflect differently from metal materials, and corner reflectors formed by the intersection of three metal surfaces are commonly used as millimeter-wave signal markers. These methods use millimeter-wave radars and cameras to simultaneously detect a certain number of stationary corner reflectors, obtaining millimeter-wave point clouds and images of the markers. These methods then use pose recovery methods such as PnP to calculate the extrinsic parameters of the millimeter-wave radar-camera.

[0003] However, existing millimeter-wave radar-camera calibration still has some problems. First, the number of angle reflections in practical applications is limited, and it is difficult to obtain sufficient observation data through static observations, so the calibration accuracy is limited by the number of observations. Second, millimeter-wave observations are easily affected by noise such as background objects and multipath signals. In restricted scenarios, such as when an unmanned boat has difficulty finding a clear background on the water, it is more difficult to extract angle reflections after mixing multiple noises. Finally, the three metal surfaces of the angle reflection are large, and the signal is reflected multiple times and returns along the original path, making it difficult to trace the position of the signal reflection point and the reflection path. Therefore, it is difficult to accurately find the pixel position of the millimeter-wave signal reflection point from the image.

[0004] Therefore, it is necessary to design a high-precision millimeter-wave radar-camera calibration method and system based on angular reflection to address the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a high-precision millimeter-wave radar-camera calibration method and system based on angle reflection to address the problems of limited number of angle reflections, sparse millimeter-wave radar point clouds, and difficulty in determining the millimeter-wave signal reflection points of angle reflections. The method increases angle reflection measurement by moving the carrier, uses GNSS to measure the GNSS coordinates of the angle reflection and the GNSS position of the carrier, calculates the distance from the angle reflection to the millimeter-wave radar at each moment based on the GNSS coordinates of the two, pre-selects the angle reflection point cloud coordinates based on the method, determines the final pixel coordinates through grid traversal and nonlinear optimization algorithm, and obtains the optimized millimeter-wave radar-camera extrinsic parameters.

[0006] According to one aspect of this specification, a high-precision millimeter-wave radar-camera calibration method based on angle reflection is provided, comprising:

[0007] The millimeter-wave radar and the camera are installed on a mobile carrier. During the movement, the millimeter-wave radar collects angular reflection point cloud data, and the camera collects angular reflection image data;

[0008] Measure and obtain the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, calculate the distance and direction from the inversion to the millimeter-wave radar based on the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, and combine the point cloud data of the inversion to obtain the point cloud coordinates of the inversion;

[0009] Based on the angular inversion image data, a grid traversal is performed starting from the angular inversion center point to obtain the preselected pixel coordinates of the millimeter wave radar reflection point;

[0010] Based on the coordinates of the angular inverse point cloud and the pixel coordinates of the angular inverse center point, a preliminary estimate of the millimeter-wave radar-camera extrinsic parameters is obtained. Combined with the pre-selected pixel coordinates of the millimeter-wave radar reflection point, a nonlinear optimization algorithm is used for iterative optimization to obtain the final millimeter-wave radar-camera extrinsic parameters and complete the calibration.

[0011] Furthermore, the angular inverse point cloud coordinates are obtained, including:

[0012] Based on the distance and direction from the angle reflection to the millimeter-wave radar, the angle reflection point cloud data is filtered by setting the distance error threshold and the angle error threshold to obtain the angle reflection point cloud coordinates.

[0013] Furthermore, obtaining the preselected pixel coordinates of the millimeter wave radar reflection point includes:

[0014] Starting from the inverse center point, grid traversal is performed, and the pixel points in the inverse image data and the inverse point cloud data are regarded as a set of point cloud pixel pairs;

[0015] Establishing an optimization function to iteratively optimize the set of point cloud pixel pairs;

[0016] The residuals of the optimization function are calculated one by one, and the point with the smallest residual is selected as the preselected pixel coordinate.

[0017] Furthermore, a nonlinear optimization algorithm is used for iterative optimization, including:

[0018] The Levenberg-Marquardt method is used to iteratively optimize the pre-selected pixel coordinates to obtain the final pixel coordinates of the millimeter-wave radar reflection point;

[0019] Based on the final pixel coordinates and angular inverse point cloud coordinates of the millimeter-wave radar reflection point, the final millimeter-wave radar-camera extrinsic parameters are obtained.

[0020] According to one aspect of this specification, a high-precision millimeter-wave radar-camera calibration system based on angular reflection is provided, comprising:

[0021] A data acquisition module is used to install the millimeter-wave radar and the camera on a mobile carrier. During the movement, the millimeter-wave radar collects angular reflection point cloud data, and the camera collects angular reflection image data;

[0022] A point cloud coordinate calculation module is used to measure and obtain the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, calculate the distance and direction from the inversion to the millimeter wave radar based on the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, and obtain the point cloud coordinates of the inversion by combining the point cloud data of the inversion;

[0023] The pixel coordinate extraction module is used to perform grid traversal starting from the center point of the angular inversion based on the angular inversion image data to obtain the pre-selected pixel coordinates of the millimeter wave radar reflection point;

[0024] The optimization and calibration module is used to obtain a preliminary estimate of the millimeter-wave radar-camera extrinsic parameters based on the coordinates of the angular inverse point cloud and the pixel coordinates of the angular inverse center point. Combined with the pre-selected pixel coordinates of the millimeter-wave radar reflection point, it uses a nonlinear optimization algorithm for iterative optimization to obtain the final millimeter-wave radar-camera extrinsic parameters and complete the calibration.

[0025] According to one aspect of the present specification, an electronic device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the high-precision millimeter-wave radar-camera calibration method based on angular reflection when executing the computer program.

[0026] 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 high-precision millimeter-wave radar-camera calibration method based on angular reflection are implemented.

[0027] According to one aspect of the present specification, a computer program product comprising instructions is provided, which, when executed on a computer, enables the computer to perform the steps of the high-precision millimeter-wave radar-camera calibration method based on angle reflection.

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

[0029] 1. The present invention installs the millimeter-wave radar and the camera on a mobile carrier, so that the millimeter-wave radar and the camera observe the angular reflection at different positions, thereby increasing the number of observations and improving the calibration accuracy.

[0030] 2. The present invention uses GNSS to measure the GNSS coordinates of the angle reflection and the GNSS position of the carrier, calculates the distance from the angle reflection to the millimeter-wave radar at each moment based on the GNSS coordinates of the two, and pre-selects the angle reflection point cloud coordinates, thereby reducing the influence of noise such as background objects and multipath signals.

[0031] 3. The present invention preselects the pixel coordinates of the reflection point by traversing each point of the angular reflection image through a grid, and uses a nonlinear optimization algorithm to determine the final pixel coordinates, thereby solving the problem of finding the pixel position of the millimeter wave signal reflection point accurately from the image. 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 following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

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

[0034] Figure 2 This is a schematic diagram of millimeter-wave radar-camera extrinsic parameters according to an embodiment of the present invention;

[0035] Figure 3 This is a schematic diagram of the millimeter wave radar-camera extrinsic parameter calibration principle according to an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the point cloud and pixel preselection principle of an embodiment of the present invention;

[0037] Figure 5 Schematic diagram of the corner inverse point cloud before and after screening according to an embodiment of the present invention;

[0038] Figure 6 Schematic diagram of the reprojection result of the angular inverse millimeter wave point cloud on the image according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] 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.

[0040] The present invention provides a high-precision millimeter-wave radar-camera calibration method based on angular reflection, such as Figure 1As shown, the method includes: installing a millimeter-wave radar and a camera on a mobile carrier, during which the millimeter-wave radar collects point cloud data of the angle inversion and the camera collects image data of the angle inversion; measuring and obtaining the GNSS coordinates of the center point of the angle inversion and the GNSS coordinates of the mobile carrier, and calculating the distance from the angle inversion to the millimeter-wave radar, and combining the point cloud data of the angle inversion to obtain the point cloud coordinates of the angle inversion; based on the angle inversion image data, performing a grid traversal starting from the center point of the angle inversion to obtain the preselected pixel coordinates of the millimeter-wave radar reflection point; based on the point cloud coordinates of the angle inversion and the pixel coordinates of the center point of the angle inversion, obtaining a preliminary estimate of the millimeter-wave radar-camera extrinsic parameters, combining the preselected pixel coordinates of the millimeter-wave radar reflection point, and using a nonlinear optimization algorithm for iterative optimization to obtain the final millimeter-wave radar-camera extrinsic parameters, thereby completing the calibration.

[0041] Specifically, suppose that the pose The pose of coordinate system a in coordinate system b is ,in Represents the rotation of the three coordinate axes, Represents the translation of the origin of the coordinate system. Assume that the millimeter wave radar-camera extrinsic parameters are , where M represents the millimeter wave radar coordinate system and C represents the camera coordinate system. Calibration is to determine The value of Figure 2 shown.

[0042] Specifically, if Figure 3 As shown, this embodiment of the present invention uses three corner reflectors placed stationary and evenly distributed in three-dimensional space. To increase the number of observations, either the corner reflectors or the carrier can be moved; there's no essential difference between the two approaches. Because manually moving the corner reflectors is labor-intensive, the present invention proposes moving the carrier during calibration to obtain a point cloud and pixel sequence from the corner reflectors. Figure 3 The number of point clouds on the yellow solid line is the number of observations, which depends on the millimeter wave sampling rate and the experiment duration.

[0043] Specifically, assuming that the coordinates of the millimeter wave point cloud of the angle reflection are , the pixel coordinates of the inverse angle are The matrix transformation equation including external parameters is as follows:

[0044] (1)

[0045] in, is the camera internal parameter, It is the millimeter wave radar-camera external parameter, is the depth of the point cloud when projected into the camera coordinate system. Expanding equation (1) yields the following:

[0046] (2)

[0047] Further expand formula (2) and write it into the form of equations as follows:

[0048] (3)

[0049] Substituting formula (3) into formula (1) and formula (2), we get:

[0050] (4)

[0051] According to formula (4), A is the coefficient matrix, F is the column vector of the external parameters, and for a set of matching points, there will be two equations. One equation will have 12 coefficients related to the external parameters, and one equation corresponds to a row in A. Then, for n sets of matching points, there will be 2n equations, 2n×12 coefficients, which can be expressed as the following equation:

[0052] (5)

[0053] Formula (5) is an overdetermined equation, that is, the number of observations far exceeds the number of unknowns. For example, if the sampling rate of the millimeter wave radar and camera is set to 2Hz, 600 matching points can be obtained after collecting for 5 minutes. The least squares method is used to calculate Matrix, you can recover the external parameters .

[0054] Specifically, although point clouds and images are obtained, the point clouds and pixel observation values ​​of the corner inversion are difficult to determine. Since millimeter-wave radar signals are easily interfered with by background objects, multipath signals and other noise, it is difficult to extract the corner inversion point cloud. On the other hand, the corner inversion is made by the intersection of three planes, and the signal is reflected multiple times and returned along the original path. The position of the reflection point on the corner inversion is difficult to determine, which makes it difficult to determine the pixel observation value of the reflection point on the image. If it is selected arbitrarily, it will inevitably introduce large errors, which will directly affect the calibration results. Therefore, it is necessary to pre-select the corner inversion point cloud and pixel observation values.

[0055] Specifically, the embodiment of the present invention also provides a method and steps for pre-selecting the angle inverse point cloud. First, the relative position relationship between the absolute position of the angle inverse and the millimeter wave radar at different observation positions is obtained by the GNSS device, the interference data in the point cloud data is filtered out, and the angle inverse point cloud data corresponding to the image data is preliminarily screened. The GNSS coordinates of the angle inverse center point are measured by the GNSS device to obtain its latitude in the global coordinate system. ,longitude and elevation At the same time, the GNSS module on the millimeter wave radar carrier is used to obtain the current position and attitude information to determine the latitude of the millimeter wave radar in the global coordinate system. ,longitude and elevation Then the GNSS coordinates are converted to Earth-centered Earth-fixed coordinates to obtain the absolute position of the millimeter-wave radar in Earth-centered Earth-fixed coordinates. and , omitting the absolute position subscript, the geodetic coordinates Earth-centered Earth-fixed coordinates The conversion relationship between them is:

[0056] (6)

[0057] in, For latitude, is the longitude, is the elevation, , is the meridian curvature radius, middle, is the first eccentricity of the Earth ellipsoid, is the semi-major axis of the Earth ellipsoid. and the geodetic coordinates of millimeter-wave radar After substituting them into the above formula, we can get the inverse Earth-centered Earth-fixed coordinates and the Earth-centered Earth-fixed coordinates of millimeter-wave radar .

[0058] After obtaining the two absolute positions at each moment, the relative position between the millimeter-wave radar and the angle inverse is calculated, and the coordinates of the angle inverse relative to the millimeter-wave radar can be obtained as follows: , recorded as In point cloud data, the inverse point cloud coordinates are , using the distance and direction information of millimeter wave radar and angle reflection, by setting the distance error threshold and angle error threshold To filter, the pre-selected point cloud must meet the following conditions:

[0059] (7)

[0060] The distance error threshold and angle error threshold limit the position of the angle reflection in the point cloud to the position obtained by GNSS. Points that meet the conditions are used as pre-selected points to effectively remove noise interference. After filtering and pre-selection, the millimeter-wave radar point cloud can be obtained as a point cloud with unique angle reflection.

[0061] Specifically, the embodiment of the present invention also provides a method step for preselecting the pixel coordinates of the reflection point. The correlation point of millimeter wave radar Transform to the camera coordinate system, and then use the camera's internal parameters It is further transformed to the image plane. Calculate the associated points of the millimeter wave radar The transformed image plane position and the preselected pixel points in the image The Euclidean distance, which is the reprojection error function, is used as a nonlinear optimization function as follows:

[0062] (8)

[0063] in, It is the serial number of the angle. is the number of preselected pixels to match a single point cloud, For external reference .

[0064] Specifically, if Figure 4 As shown, the embodiment of the present invention provides a principle of point cloud and pixel preselection. First, starting from the center point of the corner in the image, the corner is preselected according to the preset pixel interval. , which sets a grid with fixed side lengths. The selection starts at the center of the inverse triangle and gradually expands outward along the grid. These pre-selected pixels are each associated with their corresponding point cloud and used as a set of point cloud pixel pairs for iterative optimization. This method is repeated, calculating the residual of the optimization function one by one, and selecting the point with the smallest residual as the pre-selected pixel.

[0065] Specifically, if there are no obvious obstacles near the corner inversion, the corner inversion point cloud obtained after point cloud preselection is usually a true corner inversion point cloud observation. However, the grid method only roughly preselects the pixel coordinates of the reflection point, which still differs from the true pixel observation value. To further improve the estimation accuracy, the exact pixel coordinates of the reflection point need to be iteratively optimized after the point cloud preselection. By iteratively estimating both the reflection point pixel and the millimeter-wave radar-camera extrinsic parameters to be estimated as optimization parameters, the final extrinsic parameter optimization estimation is completed while continuously reselecting the reflection point pixel coordinates.

[0066] The embodiment of the present invention also provides a method and steps for the joint optimization of the angle inversion pixel coordinates / external parameters. With the preselected pixel coordinates as the center, the final reflection point pixel coordinates are restricted to the pixel area where the angle inversion is located, and a constrained Lagrangian function is constructed. The reflection point pixel coordinates are , the coordinates of the three vertices of the corner reflector in the image are , the constraints are as follows:

[0067] (9)

[0068] The constraints for the same angle inversion are the same, and the optimization function is:

[0069] (10)

[0070] The Lagrangian function with constraints is:

[0071] (11)

[0072] in, is the Lagrange multiplier. The above formula contains the optimization function The LM algorithm (Levenberg-Marquardt algorithm) is used to iteratively optimize pixel observations and extrinsic parameters using three Lagrangian functions. The LM algorithm combines the advantages of gradient descent and Gauss-Newton methods, adjusting the optimization direction and step size based on the error. The LM algorithm iteratively adjusts pixel observations and camera extrinsics, continuously reselecting pixel observations at reflection points within a limited range and obtaining optimized millimeter-wave radar-camera extrinsics.

[0073] Specifically, the embodiment of the present invention pre-screens the angular inverse point cloud using GNSS coordinates in a millimeter wave point cloud containing more noise in an interference scenario. Figure 5 As shown in Figure (a), it represents the original point cloud without screening, from which the points with reversed angles cannot be distinguished; Figure 5 As shown in Figure (b), the corner inverse point cloud screened out with GNSS assistance is marked as a larger square.

[0074] Specifically, this embodiment of the present invention also provides the results of millimeter-wave radar-camera extrinsic calibration, listed in Table 1. As can be seen from the table, the maximum rotation angle calibration error is 0.32°, and the maximum translation error is 0.2m. The millimeter-wave radar point cloud is reprojected onto the image using the extrinsic parameters, and the reprojection error is calculated. The average reprojection error for eight pairs of points is 3.857 pixels.

[0075] Table 1 Millimeter-wave radar-camera calibration results

[0076]

[0077] like Figure 6 As shown, the embodiment of the present invention also provides the reprojection result of the angular inverse millimeter wave point cloud on the image. Figure 6 The red point is the effect of projecting the 3D point onto the image. It can be seen that the reprojection result is basically within the range of the angular inverse pixel.

[0078] The implementation of each embodiment of the present invention is based on programmed processing performed by a device with processor functionality. Therefore, in practical engineering, the technical solutions and functionalities of each embodiment of the present invention are packaged into various modules. Based on this reality, and in addition to the aforementioned embodiments, an embodiment of the present invention provides a high-precision millimeter-wave radar-camera calibration system based on angular inversion. This system is used to implement the high-precision millimeter-wave radar-camera calibration method based on angular inversion described in the aforementioned method embodiment.

[0079] The system includes: a data acquisition module, which is used to install the millimeter-wave radar and the camera on a mobile carrier. During the movement, the millimeter-wave radar collects point cloud data of the angle reflection, and the camera collects image data of the angle reflection; a point cloud coordinate calculation module, which is used to measure the GNSS coordinates of the center point of the angle reflection and the GNSS coordinates of the mobile carrier, and calculate the distance from the angle reflection to the millimeter-wave radar, and obtain the point cloud coordinates of the angle reflection in combination with the point cloud data of the angle reflection; a pixel coordinate extraction module, which is used to perform grid traversal starting from the center point of the angle reflection based on the angle reflection image data to obtain the pre-selected pixel coordinates of the millimeter-wave radar reflection point; an optimization and calibration module, which is used to obtain a preliminary estimate of the millimeter-wave radar-camera extrinsic parameters based on the point cloud coordinates of the angle reflection and the pixel coordinates of the center point of the angle reflection, and use a nonlinear optimization algorithm to perform iterative optimization in combination with the pre-selected pixel coordinates of the millimeter-wave radar reflection point to obtain the final millimeter-wave radar-camera extrinsic parameters and complete the calibration.

[0080] An embodiment of the present invention provides a high-precision millimeter-wave radar-camera calibration system based on angle reflection, which solves the problems of a limited number of angle reflections, a sparse millimeter-wave radar point cloud, and difficulty in determining the millimeter-wave signal reflection point of the angle reflection. By adopting the above-mentioned several modules, by installing the millimeter-wave radar and the camera on a mobile carrier, the millimeter-wave radar and the camera observe the angle reflection at different positions, the number of observations is increased, and the calibration accuracy is improved. By using GNSS to measure the GNSS coordinates of the angle reflection and the GNSS position of the carrier, the distance from the angle reflection to the millimeter-wave radar at each moment is calculated based on the GNSS coordinates of the two and the coordinates of the angle reflection point cloud are pre-selected, thereby reducing the influence of noise such as background objects and multipath signals. The pixel coordinates of the reflection point are pre-selected by traversing each point of the angle reflection image through a grid, and the final pixel coordinates are determined by a nonlinear optimization algorithm, thereby solving the problem of difficulty in accurately finding the pixel position of the millimeter-wave signal reflection point from the image.

[0081] 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, wherein the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the high-precision millimeter-wave radar-camera calibration method based on angular reflection as proposed in the aforementioned embodiment.

[0082] An embodiment of the present invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, this program is used to reduce the effects of noise such as background objects and multipath signals, increase the number of observations, and improve calibration accuracy. The storage medium can be any non-volatile storage device, such as a hard disk, solid-state drive, flash drive, or optical disk, and is used to store the computer program code and necessary data files. The stored computer program includes a data acquisition module, a point cloud coordinate calculation module, a pixel coordinate extraction module, and an optimization and calibration module.

[0083] Embodiments of the present invention also provide a computer program product containing instructions that, when executed on a computer, fully or partially implements the high-precision millimeter-wave radar-camera calibration method based on angular reflection according to the above embodiments. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0084] Finally, it should be noted that the above specific embodiments are merely representative examples of the present invention. Obviously, the present invention is not limited to the above specific embodiments and is susceptible to numerous variations. Any simple modifications, equivalent variations, and modifications to the above specific embodiments based on the technical essence of the present invention shall be deemed to fall within the scope of protection of the present invention.

Claims

1. A high-precision millimeter-wave radar-camera calibration method based on angular reflection, characterized in that: By matching the inverse point cloud and pixels, the millimeter-wave radar-camera extrinsic parameters are directly solved, including: The millimeter-wave radar and the camera are installed on a mobile carrier. During the movement, the millimeter-wave radar collects angular reflection point cloud data, and the camera collects angular reflection image data; Measure and obtain the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, calculate the distance and direction from the inversion to the millimeter-wave radar based on the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, and combine the point cloud data of the inversion to obtain the point cloud coordinates of the inversion; Based on the image data of the angle reflection, the grid is traversed starting from the center point of the angle reflection to obtain the pre-selected pixel coordinates of the millimeter wave radar reflection point; Based on the angular inverse point cloud coordinates and the pre-selected pixel coordinates of the millimeter-wave radar reflection point, a preliminary estimate of the millimeter-wave radar-camera extrinsic parameters is obtained. The pre-selected pixel coordinates are iteratively optimized using the Levenberg-Marquardt method to obtain the final pixel coordinates of the millimeter-wave radar reflection point. Based on the final pixel coordinates of the millimeter-wave radar reflection point and the angular inverse point cloud coordinates, the final millimeter-wave radar-camera extrinsic parameters are obtained to complete the calibration.

2. The high-precision millimeter-wave radar-camera calibration method based on angular reflection according to claim 1, characterized in that: Get the inverse point cloud coordinates, including: Based on the distance and direction from the angle reflection to the millimeter-wave radar, the angle reflection point cloud data is filtered by setting the distance error threshold and the angle error threshold to obtain the angle reflection point cloud coordinates.

3. The high-precision millimeter-wave radar-camera calibration method based on angular reflection according to claim 1, characterized in that: Get the preselected pixel coordinates of the millimeter wave radar reflection point, including: Starting from the inverse center point, grid traversal is performed, and the pixel points in the inverse image data and the inverse point cloud data are regarded as a set of point cloud pixel pairs; Establishing an optimization function to iteratively optimize the set of point cloud pixel pairs; The residuals of the optimization function are calculated one by one, and the point with the smallest residual is selected as the preselected pixel coordinate.

4. A high-precision millimeter-wave radar-camera calibration system based on angular reflection, characterized in that: include: A data acquisition module is used to install the millimeter-wave radar and the camera on a mobile carrier. During the movement, the millimeter-wave radar collects angular reflection point cloud data, and the camera collects angular reflection image data; A point cloud coordinate calculation module is used to measure and obtain the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, calculate the distance and direction from the inversion to the millimeter wave radar based on the GNSS coordinates of the center point of the inversion and the GNSS coordinates of the mobile carrier, and obtain the point cloud coordinates of the inversion by combining the point cloud data of the inversion; The pixel coordinate extraction module is used to perform grid traversal starting from the center point of the angular inversion based on the angular inversion image data to obtain the pre-selected pixel coordinates of the millimeter wave radar reflection point; The optimization and calibration module is used to obtain a preliminary estimate of the millimeter-wave radar-camera extrinsic parameters based on the angular inverse point cloud coordinates and the pixel coordinates of the angular inverse center point. The preselected pixel coordinates are iteratively optimized using the Levenberg-Marquardt method to obtain the final pixel coordinates of the millimeter-wave radar reflection point. The final millimeter-wave radar-camera extrinsic parameters are obtained based on the final pixel coordinates of the millimeter-wave radar reflection point and the angular inverse point cloud coordinates, completing the calibration.

5. 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 high-precision millimeter-wave radar-camera calibration method based on angle reflection according to any one of claims 1 to 3 are implemented.

6. 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 high-precision millimeter-wave radar-camera calibration method based on angle reflection according to any one of claims 1 to 3 are implemented.

7. A computer program product comprising instructions, characterized in that When the method is run on a computer, the computer is enabled to execute the steps of the high-precision millimeter-wave radar-camera calibration method based on angle reflection according to any one of claims 1 to 3.

Images