AGV Bottom Camera and LiDAR Joint Calibration Method, System and Device

Abstract

The present invention discloses a joint calibration method, system and device based on an AGV bottom camera and a lidar, which relates to the technical field of multi-sensor data fusion. The method includes: obtaining the pose of the bottom camera and the pose of the first lidar; aligning the pose of the first lidar and the pose of the bottom camera in the mapping mode to construct a prior map; obtaining N poses of the second lidar by matching the prior map, and simultaneously converting the N poses of the second lidar into the pose of the bottom camera in the positioning mode to calibrate the pose of the bottom camera and the pose of the second lidar. The main technical solutions and effects are as follows: in the positioning mode, a high-precision prior map is obtained through the mapping mode and manual modification, and at the same time, the poses of the second lidar are quickly obtained. The optimal coordinate transformation parameters are determined by combining the minimum residual sum with the Ceres iterative algorithm, and the alignment of the poses of the second lidar to the pose of the bottom camera is quickly completed, realizing the simultaneous, fast and high-precision calibration of multiple AGV cars.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-sensor data fusion, and particularly to a joint calibration method, system and device based on an AGV bottom camera and a lidar. Background Art

[0002] To solve the problem of a robot performing real-time positioning and mapping in an unknown environment, the Simultaneous Localization and Mapping (SLAM) technology has emerged and become the core challenge in the AGV technology system. The quality of the SLAM effect is closely related to the quality of sensor data. To obtain better accuracy and robustness, multiple sensor fusion methods are usually introduced to enrich the input information. Therefore, the joint calibration between sensors ensures the quality of the fused information.

[0003] In the existing AGV technology, lidar and a bottom camera are usually used for environmental perception, and mapping and positioning are performed in combination with the SLAM algorithm. The lidar is responsible for perceiving the surrounding environment, and the bottom camera is responsible for scanning the QR code on the ground to obtain its own position. The AGV mainly uses the bottom camera as the coordinate center. Therefore, the laser information of the lidar needs to be unified into the AGV coordinate system. At this time, the conversion relationship from the lidar to the bottom camera is required. The existing joint calibration scheme requires many dependent wrapper programs and precise calibration plates, with high costs and a cumbersome process. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, a simple and low-cost technical solution is needed to achieve the joint calibration of the lidar and the bottom camera in AGV technology. The present invention discloses a joint calibration method, system and device based on an AGV bottom camera and a lidar.

[0005] A joint calibration method based on an AGV bottom camera and a lidar includes the following steps:

[0006] Obtain the pose of the bottom camera and the first lidar pose;

[0007] Match the pose of the bottom camera and the first lidar pose with the same timestamp to form N pairs of first matching pose pairs;

[0008] In the mapping mode, construct a prior map by aligning the first lidar pose and the bottom camera pose in the first matching pose pairs;

[0009] Obtain N second lidar poses by matching the prior map, and match the second lidar poses with the pose of the bottom camera with the same timestamp to obtain N pairs of second matching pose pairs;

[0010] In the positioning mode, simultaneously convert N second radar poses in the second matching pose pair into the bottom camera pose to calibrate the bottom camera pose and the second radar pose.

[0011] As an implementable manner, the obtaining of the bottom camera pose includes the following steps:

[0012] Obtain a QR code label and a QR code image through the AGV bottom camera, and obtain the QR code pose by identifying the QR code label;

[0013] Obtain the pose of the QR code in the QR code image through deep learning methods combined with image processing, and obtain the bottom camera pose through calculation;

[0014] The bottom camera pose is expressed as follows:

[0015]

[0016] Where, represents the bottom camera pose, represents the pose of the QR code in the QR code image, represents the QR code pose.

[0017] As an implementable manner, the obtaining of the first radar pose includes the following steps:

[0018] Take the initialized camera pose as the initial radar pose, and create a new map with the initial radar pose as the first frame of the map;

[0019] Obtain the relative pose of the current frame to the next frame of radar through the ICP algorithm, and obtain the next frame of radar pose through calculation;

[0020] The specific calculation formula for obtaining the next frame of radar pose through the ICP algorithm is:

[0021]

[0022] Where, represents the next frame of radar pose, represents the current frame of radar pose, represents the relative pose of the current frame to the next frame of radar.

[0023] As an implementable manner, the alignment of the first radar pose and the bottom camera pose in the mapping mode includes the following steps:

[0024] Perform a coordinate transformation on the first radar pose so that the first radar pose and the camera pose are in the same world coordinate system, and obtain the radar pose in the world coordinate system. The coordinate transformation matrix is denoted as the first transformation matrix;

[0025] Convert the radar pose in the world coordinate system into the camera pose, and denote the conversion matrix as the second transformation matrix;

[0026] Determine the first transformation matrix and the second transformation matrix through the minimum residual sum;

[0027] Further obtain the transformation parameters through the Ceres iterative algorithm, so as to align the bottom camera pose and the first radar pose.

[0028] As an implementable manner, the calculation formula of the minimum residual sum is:

[0029]

[0030] where represents the minimum residual sum, represents the i-th residual in N first matching pose pairs;

[0031] The calculation formula of the residual is:

[0032]

[0033] where represents the i-th residual in N first matching pose pairs, represents the bottom camera pose, represents the estimated pose of the bottom camera;

[0034] The calculation formula of the estimated pose of the bottom camera is:

[0035]

[0036] where represents the estimated pose of the bottom camera, represents the second transformation matrix, represents the first transformation matrix, represents the first radar pose.

[0037] As an implementable manner, obtaining N second radar poses by matching the prior map includes the following steps:

[0038] Collect the matching points of N radars to obtain N radar trajectories;

[0039] The N second radar poses are obtained from the N radar trajectories by matching the prior map.

[0040] As an implementable manner, converting the second radar pose in the several second matching pose pairs into the bottom camera pose in the positioning mode includes the following steps:

[0041] Convert the second radar pose coordinates to the camera pose, and denote the conversion matrix as the third transformation matrix;

[0042] Determine the third transformation matrix by minimizing the sum of residuals;

[0043] Further obtain the transformation parameters through the Ceres iterative algorithm, so as to align the bottom camera pose and the second radar pose.

[0044] As an implementable manner, the conversion of the second radar pose coordinates to the camera pose is expressed as follows:

[0045]

[0046] Where, represents the predicted value of the third transformation matrix, represents the second radar pose, represents the bottom camera pose;

[0047] The sum of the minimum residuals is expressed as follows:

[0048]

[0049] Where, represents the third transformation matrix, represents the predicted value of the third transformation matrix.

[0050] As an implementable manner, the prior map is manually modified to obtain a high-precision prior map.

[0051] A joint calibration system based on an AGV bottom camera and a lidar includes: a pose acquisition module, a first matching pose pair module, a prior map construction module, a second matching pose pair module, and a positioning calibration module;

[0052] The pose acquisition module is used to acquire the bottom camera pose and the first radar pose;

[0053] The first matching pose pair module is used to match the bottom camera pose and the first radar pose with the same timestamp to form N pairs of first matching pose pairs;

[0054] The prior map construction module is used to align the first radar pose and the bottom camera pose in the first matching pose pair in the mapping mode to construct a prior map;

[0055] The second matching pose pair module is used to obtain N second radar poses by matching the prior map, and match the second radar poses with the bottom camera poses with the same timestamp to obtain N pairs of second matching pose pairs;

[0056] The positioning and calibration module is used to simultaneously convert N second radar poses in the second matching pose pair into the bottom camera pose in the positioning mode, so as to calibrate the bottom camera pose and the second radar pose.

[0057] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following method is implemented:

[0058] Obtain the bottom camera pose and the first radar pose;

[0059] Match the bottom camera pose and the first radar pose with the same time stamp to form N pairs of first matching pose pairs;

[0060] In the mapping mode, align the first radar pose and the bottom camera pose in the first matching pose pair to construct a prior map;

[0061] Obtain N second radar poses by matching the prior map, and match the second radar poses with the bottom camera poses with the same time stamp to obtain N pairs of second matching pose pairs;

[0062] In the positioning mode, simultaneously convert N second radar poses in the second matching pose pair into the bottom camera pose, so as to calibrate the bottom camera pose and the second radar pose.

[0063] A joint calibration device based on an AGV bottom camera and a lidar includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the following method is implemented:

[0064] Obtain the bottom camera pose and the first radar pose;

[0065] Match the bottom camera pose and the first radar pose with the same time stamp to form N pairs of first matching pose pairs;

[0066] In the mapping mode, align the first radar pose and the bottom camera pose in the first matching pose pair to construct a prior map;

[0067] Obtain N second radar poses by matching the prior map, and match the second radar poses with the bottom camera poses with the same time stamp to obtain N pairs of second matching pose pairs;

[0068] In the positioning mode, simultaneously convert N second radar poses in the second matching pose pair into the bottom camera pose, so as to calibrate the bottom camera pose and the second radar pose.

[0069] Due to the adoption of the above technical solutions, the present invention has remarkable technical effects:

[0070] 1. In the mapping mode of the present invention, the optimal coordinate transformation parameters are determined by combining the minimum residual sum with the Ceres iterative algorithm, so as to align the pose of the first radar with the pose of the bottom camera. It realizes that there is no site limitation and no need for specific calibration instruments in the mapping mode, and the AGV camera-lidar calibration result can be obtained simply and quickly, while greatly reducing the cost.

[0071] 2. In the positioning mode of the present invention, a high-precision prior map is obtained through the mapping mode and manual modification. The pose of the second radar can be quickly obtained at the same time, and the optimal coordinate transformation parameters are determined by combining the minimum residual sum with the Ceres iterative algorithm, and the alignment of the pose of the second radar with the pose of the bottom camera is quickly completed, realizing the simultaneous rapid and high-precision calibration of multiple AGV cars. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a schematic flowchart of the joint calibration method based on the bottom camera and lidar of the AGV described in the present invention;

[0073] Figure 2 It is an overall schematic diagram of the joint calibration system based on the bottom camera and lidar of the AGV described in the present invention;

[0074] Figure 3 It is a diagram of the bottom camera trajectory and radar trajectory in the mapping mode of the embodiment of the present invention;

[0075] Figure 4 It is a diagram of the bottom camera trajectory and radar trajectory in the positioning mode of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0076] The present invention will be further described below in conjunction with the drawings and embodiments. The following embodiments are explanations of the present invention and the present invention is not limited to the following embodiments.

[0077] Embodiment 1:

[0078] A joint calibration method based on the bottom camera and lidar of an AGV, as Figure 1 shown, includes the following steps:

[0079] S100. Obtain the pose of the bottom camera and the pose of the first radar;

[0080] S200. Match the pose of the bottom camera and the pose of the first radar with the same time stamp to form N pairs of first matching pose pairs;

[0081] S300. In the mapping mode, construct a prior map by aligning the pose of the first radar and the pose of the bottom camera in the first matching pose pair;

[0082] S400. Obtain N second radar poses by matching the prior map, and match the second radar poses with the bottom camera poses at the same timestamp to obtain N pairs of second matching pose pairs;

[0083] S500. In the positioning mode, simultaneously convert the N second radar poses in the second matching pose pairs into bottom camera poses to calibrate the bottom camera poses and the second radar poses.

[0084] In the mapping mode of the present invention, the bottom camera pose and the radar pose are obtained, and the first radar pose at the same time is converted into the camera pose through coordinate transformation. The minimum residual sum and the Ceres iterative algorithm are combined to determine the optimal coordinate transformation parameters, realizing the precise alignment of the first radar pose and the bottom camera pose, and obtaining the prior map; in the positioning mode, the bottom camera pose is obtained and multiple second radar poses are simultaneously obtained by matching the prior map. The minimum residual sum and the Ceres iterative algorithm are combined to determine the optimal coordinate transformation parameters, quickly completing the alignment of the second radar pose to the bottom camera pose, and realizing the simultaneous, fast and high-precision calibration of multiple AGV cars.

[0085] In this embodiment, in step S100, obtaining the bottom camera pose and the first radar pose specifically includes the following steps:

[0086] S110: Use the AGV bottom camera to obtain the QR code label and the QR code image, and obtain the QR code pose by identifying the QR code label (this step realizes the conversion of the world coordinate system to the QR code coordinate system);

[0087] S120: Obtain the pose of the QR code in the QR code image through deep learning method combined with image processing (this step realizes the conversion of the QR code coordinate system to the camera coordinate system), and obtain the bottom camera pose through calculation.

[0088] The bottom camera pose is expressed as follows:

[0089]

[0090] Among them, represents the bottom camera pose, represents the pose of the QR code in the image, represents the QR code pose.

[0091] S130: Take the initialized camera pose as the initial radar pose, and create a new map with this initial radar pose as the first frame of the map;

[0092] S140: Obtain the relative pose of the current frame to the next frame of radar through the ICP algorithm, obtain the next frame of radar pose through calculation, and insert this pose into the map.

[0093] The specific calculation formula for obtaining the next-frame radar pose through the ICP algorithm is as follows:

[0094]

[0095] where represents the next-frame radar pose, represents the initial pose, represents the relative pose from the current frame to the next-frame radar.

[0096] In this embodiment, in step S200, the bottom camera pose and the first radar pose with the same timestamp are matched to form N pairs of first matching pose pairs. Specifically, the structure of the AGV is relatively fixed, and the relative positions of the lidar and the bottom camera are also basically fixed. Therefore, during the AGV's travel, their trajectories are relatively parallel. The trajectories of the bottom camera and the radar in the mapping mode are as shown in Figure 3 In the figure, the yellow line is the first radar trajectory, and the blue line is the bottom camera trajectory. The blue line is not a single point. After magnification, it is as shown on the right in Figure 3 As can be seen from Figure 3 the bottom camera trajectory intersects the first radar trajectory at one point. This is because in the mapping mode, when the lidar scans the laser points, there is no map available for matching for initialization. Therefore, the bottom camera pose is used as the radar initial pose during initialization.

[0097] Since in this embodiment, the AGV establishes its own coordinate system centered on the bottom camera, the sensor data obtained by the radar needs to be converted from the radar pose to the AGV's own pose (bottom camera pose).

[0098] In step S300, in the mapping mode, by aligning the first radar pose and the bottom camera pose in the first matching pose pair, a prior map is constructed, which specifically includes the following steps:

[0099] S310: Perform coordinate transformation on the first radar pose so that the first radar pose and the camera pose are in the same world coordinate system, and obtain the radar pose in the world coordinate system. The coordinate transformation matrix is denoted as the first transformation matrix;

[0100] S320: Convert the radar pose in the world coordinate system to the camera pose. The conversion matrix is denoted as the second transformation matrix;

[0101] S330: Determine more accurate first and second transformation matrices through the minimum residual sum, and further optimize the conversion relationship from the first radar pose to the camera pose;

[0102] S340: Further iterate through the Ceres algorithm to obtain the optimal transformation parameters, thereby achieving the precise alignment of the bottom camera pose and the first radar pose, and further achieving calibration in the mapping mode and constructing a prior map. It realizes that there are no site restrictions and no specific calibration instruments required in the mapping mode, and can simply and quickly obtain the AGV camera-laser calibration result, while greatly reducing the cost.

[0103] The calculation formula for the sum of the minimum residuals is as follows:

[0104]

[0105] Where, represents the i-th residual in N pairs of first matching poses;

[0106] The calculation formula for the residual is as follows:

[0107]

[0108] Where, represents the i-th residual in N pairs of first matching poses, represents the bottom camera pose, represents the estimated pose of the bottom camera;

[0109] The calculation formula for the estimated pose of the bottom camera is as follows:

[0110]

[0111] Where, represents the estimated pose of the bottom camera, represents the second transformation matrix, represents the first transformation matrix, represents the first radar pose.

[0112] In this embodiment, step S400 obtains N second radar poses by matching the prior map, and matches the second radar poses with the bottom camera poses with the same timestamp to obtain N pairs of second matching pose pairs, including the following steps:

[0113] S410: Collect N pairs of matching points of multiple radars (multiple AGV cars) in the positioning mode, as Figure 4 shown, in this embodiment, the front radar trajectory and the rear radar trajectory are collected, Figure 4 where the blue is the bottom camera trajectory, the yellow is the front radar trajectory, and the green is the rear radar trajectory;

[0114] S420: The front radar trajectory and the rear radar trajectory of N pairs of matching points obtain N second radar poses by matching the prior map;

[0115] S430: Match the second radar pose with the bottom camera pose at the same timestamp to obtain N pairs of second matching pose pairs.

[0116] In this embodiment, in the positioning mode, step S500 simultaneously converts the N second radar poses in the second matching pose pairs into bottom camera poses to calibrate the bottom camera pose and the second radar pose. The specific steps are as follows:

[0117] S510: Multiple second radar poses can be obtained simultaneously by matching the prior map. The prior map can be manually modified to obtain a high-precision prior map.

[0118] S520: Convert the coordinates of the multiple second radar poses into camera poses, and the conversion matrix is denoted as the third transformation matrix.

[0119] S530: Determine a more accurate third transformation matrix through the minimum residual sum, and then optimize the conversion relationship from the radar pose to the camera pose.

[0120] S540: Further iterate through the Ceres algorithm to obtain the optimal coordinate transformation parameters, so as to simultaneously and more accurately align multiple second radar poses with the bottom camera pose, complete the calibration, and achieve fast and high-precision calibration of multiple AGV vehicles at the same time.

[0121] The conversion of the second radar pose coordinates into camera poses is expressed as follows:

[0122]

[0123] Among them, represents the predicted value of the third transformation matrix, represents the second radar pose, represents the bottom camera pose;

[0124] The minimum residual sum is expressed as follows:

[0125]

[0126] Among them, represents the third transformation matrix, represents the predicted value of the third transformation matrix.

[0127] In the mapping mode of the embodiments of the present invention, the optimal coordinate transformation parameters are determined by combining the minimum residual sum with the Ceres iterative algorithm, so as to align the pose of the first radar with the pose of the bottom camera, achieving that there is no site limitation in the mapping mode, no need for specific calibration instruments, and obtaining the AGV camera-lidar calibration result simply and quickly, while greatly reducing the cost. At the same time, in the positioning mode of the embodiments of the present invention, a high-precision prior map is obtained through the mapping mode and manual modification, the pose of the second radar can be obtained quickly at the same time, and the optimal coordinate transformation parameters are determined by combining the minimum residual sum with the Ceres iterative algorithm, and the alignment of the pose of the second radar with the pose of the bottom camera is quickly completed, realizing the simultaneous, fast and high-precision calibration of multiple AGV vehicles.

[0128] Embodiment 2:

[0129] A joint calibration system based on an AGV bottom camera and a lidar, as Figure 2 shown, includes: a pose acquisition module 100, a first matching pose pair module 200, a prior map construction module 300, a second matching pose pair module 400, and a positioning calibration module 500;

[0130] The pose acquisition module 100 is used to acquire the pose of the bottom camera and the pose of the first radar;

[0131] The first matching pose pair module 200 is used to match the pose of the bottom camera and the pose of the first radar with the same time stamp, and form N pairs of first matching pose pairs;

[0132] The prior map construction module 300 is used to construct a prior map by aligning the pose of the first radar and the pose of the bottom camera in the first matching pose pairs in the mapping mode;

[0133] The second matching pose pair module 400 is used to obtain N poses of the second radar by matching the prior map, and match the poses of the second radar with the pose of the bottom camera with the same time stamp to obtain N pairs of second matching pose pairs;

[0134] The positioning calibration module 500 is used to simultaneously convert the N poses of the second radar in the second matching pose pairs into the pose of the bottom camera in the positioning mode, so as to calibrate the pose of the bottom camera and the pose of the second radar.

[0135] All equivalent technical solutions also belong to the scope of the present invention for various changes and modifications made without departing from the spirit and scope of the present invention.

[0136] Each embodiment in this specification is described in a progressive manner, and the key points of each embodiment are the differences from other embodiments. For the same and similar parts between each embodiment, reference can be made to each other.

[0137] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, apparatus, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0138] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0139] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0140] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0141] It should be noted that:

[0142] References in this specification to "one embodiment" or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Therefore, appearances of the phrases "one embodiment" or "an embodiment" in various places throughout this specification do not necessarily refer to the same embodiment.

[0143] In addition, it should be noted that for the specific embodiments described in this specification, the shapes and names of the components can be different. Any equivalent or simple changes made according to the structure, features, and principles described in the inventive concept of this invention patent are included within the protection scope of this invention patent. Those skilled in the art of this invention can make various modifications, supplements, or use similar methods for substitution to the specific embodiments described, as long as they do not deviate from the structure of this invention or exceed the scope defined by this claims, they shall fall within the protection scope of this invention.

Claims

1. A joint calibration method based on AGV bottom camera and laser radar, characterized in that: The following steps are involved: Get the bottom camera pose and the first radar pose; Matching the bottom camera pose and the first radar pose at the same timestamp to form N pairs of first matching pose pairs; In a mapping mode, constructing a priori map by aligning a first radar pose and a bottom camera pose in the first matching pose pair; Obtain N second radar poses by matching the prior map, and match the second radar poses with the bottom camera poses at the same timestamp to obtain N pairs of second matching poses; In the positioning mode, simultaneously converting N second radar poses in the second matching pose pair into base camera poses to calibrate the base camera pose and the second radar pose; Obtaining the first radar pose includes the following steps: Initialize the camera pose as the initial radar pose, and create a new map with the initial radar pose as the first frame of the map; The relative position of the radar from the current frame to the next frame is obtained through the ICP algorithm, and the radar position of the next frame is obtained through calculation; Obtaining N second radar poses by matching the prior map includes the following steps: collecting matching points of the N radars to obtain N radar tracks; The radar trajectory obtains N second radar poses by matching the prior map; In the positioning mode, simultaneously converting the second radar poses in the plurality of second matching pose pairs into base camera poses includes the following steps: Convert the second radar pose coordinates into a camera pose, and record the conversion matrix as a third transformation matrix; determine the third transformation matrix by minimum residual sum; The transformation parameters are further obtained through the Ceres iterative algorithm to achieve alignment between the bottom camera pose and the second radar pose.

2. The joint calibration method based on AGV bottom camera and laser radar according to claim 1 is characterized in that: The step of obtaining the bottom camera pose comprises the following steps: Acquire the QR code label and QR code image through the AGV bottom camera, and obtain the QR code posture by identifying the QR code label; The position of the QR code in the QR code image is obtained by combining a deep learning method with image processing, and the position of the bottom camera is obtained by calculation; The bottom camera pose is expressed as follows: T WC =T CQR -1 *T WQR Among them, T WC represents the bottom camera pose, T CQR represents the pose of the QR code in the QR code image, T WQR Indicates the QR code pose.

3. The joint calibration method based on AGV bottom camera and laser radar according to claim 1 is characterized in that: The specific calculation formula for obtaining the radar pose of the next frame through the ICP algorithm is: T L' =H LL' *T L Among them, T L' Indicates the radar pose of the next frame, T L Indicates the radar pose of the current frame, H LL' Indicates the relative pose of the radar from the current frame to the next frame.

4. The joint calibration method based on AGV bottom camera and laser radar according to claim 1 is characterized in that: The method of aligning the first radar pose and the bottom camera pose in the first matching pose pair in the mapping mode includes the following steps: Performing coordinate transformation on the first radar pose so that the first radar pose and the camera pose are in the same world coordinate system, thereby obtaining the radar pose in the world coordinate system, wherein the coordinate transformation matrix is recorded as a first transformation matrix; Convert the radar pose in the world coordinate system to the camera pose, and record the conversion matrix as the second transformation matrix; Determining a first transformation matrix and a second transformation matrix by minimizing the residual sum; The transformation parameters are further obtained through the Ceres iterative algorithm to achieve alignment between the bottom camera pose and the first radar pose.

5. The joint calibration method based on AGV bottom camera and laser radar according to claim 4 is characterized in that: The calculation formula of the minimum residual sum is: Among them, SSR represents the minimum sum of residuals, r i represents the i-th residual in the N first matching pose pairs; The residual calculation formula is: r i =T WC -1 *T WC * Among them, r i represents the i-th residual in the N first matching pose pairs, T WC represents the bottom camera pose, T WC * represents the estimated pose of the bottom camera; The calculation formula for the estimated pose of the bottom camera is: T WC * =T L'C *T LL' *T L Among them, T WC * represents the estimated pose of the bottom camera, T L'C represents the second transformation matrix, T LL' represents the first transformation matrix, T L Indicates the first radar pose.

6. The joint calibration method based on AGV bottom camera and laser radar according to claim 1 is characterized in that: The second radar pose coordinates are converted to camera pose, which is expressed as follows: T LC * =T M -1 *T WC Among them, T LC * Represents the predicted value of the third transformation matrix, T M represents the second radar pose, T WC represents the bottom camera pose; The minimum residual sum is expressed as follows: Among them, T LC represents the third transformation matrix, T LC * Represents the predicted value of the third transformation matrix.

7. The joint calibration method based on AGV bottom camera and laser radar according to claim 1 is characterized in that: The priori map is manually modified to obtain a high-precision priori map.

8. A joint calibration system based on AGV bottom camera and laser radar, characterized in that: include: Pose acquisition module, first matching pose pair module, prior map construction module, second matching pose pair module, positioning calibration module, first radar pose acquisition module, bottom camera pose conversion module; The posture acquisition module is used to obtain the posture of the bottom camera and the first radar; The first matching pose pair module is configured to match the bottom camera pose and the first radar pose with the same timestamp to form N pairs of first matching pose pairs; The prior map construction module is configured to construct a prior map by aligning the first radar pose and the bottom camera pose in the first matching pose pair in a mapping mode; The second matching pose pair module is configured to obtain N second radar poses by matching the prior map, and match the second radar poses with the bottom camera poses at the same timestamp to obtain N pairs of second matching pose pairs; The positioning and calibration module is configured to simultaneously convert N second radar poses in the second matching pose pair into base camera poses in a positioning mode, so as to calibrate the base camera pose and the second radar pose; The first radar pose acquisition module is used to acquire the first radar pose, including the following steps: Initialize the camera pose as the initial radar pose, and create a new map with the initial radar pose as the first frame of the map; The relative position of the radar from the current frame to the next frame is obtained through the ICP algorithm, and the radar position of the next frame is obtained through calculation; Obtaining N second radar poses by matching the prior map includes the following steps: collecting matching points of the N radars to obtain N radar tracks; The radar trajectory obtains N second radar poses by matching the prior map; The bottom camera pose conversion module is used to simultaneously convert the second radar poses in the plurality of second matching pose pairs into bottom camera poses in the positioning mode, comprising the following steps: Convert the second radar pose coordinates into a camera pose, and record the conversion matrix as a third transformation matrix; determine the third transformation matrix by minimum residual sum; The transformation parameters are further obtained through the Ceres iterative algorithm to achieve alignment between the bottom camera pose and the second radar pose.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. A device based on joint calibration of an AGV bottom camera and a laser radar, 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, the method according to any one of claims 1 to 7 is implemented.

Images