Multi-sensor external parameter calibration system and method based on mechanical arm and adjustable sliding rail

Abstract

The invention relates to the technical field of sensor calibration, and discloses a multi-sensor external parameter calibration system and method based on a mechanical arm and an adjustable sliding rail, and a calibration target object unit comprises a calibration rack and a target object fixed on the calibration rack; the mechanical arm sliding rail unit comprises a mechanical arm, and a sliding rail with scales is arranged at the tail end of the mechanical arm; the mechanical arm is controlled to drive the sliding rail to a plurality of poses; the sensor unit comprises at least two sensors which are connected through a sliding rail, the relative positions of the sensors are adjusted along the sliding rail, the sensors are locked at preset scales so as to determine preset external reference values, and the sensors collect target object observation data corresponding to all poses. And the data processing unit calculates a corresponding external reference estimation value by using the observation data of the target object, optimizes to obtain a final external reference value, and compares the final external reference value with a preset external reference reference value to calculate an error. According to the invention, the external reference value is preset, the multi-view observation data is automatically acquired and fused and analyzed, and the verifiability of the calibration result and the intuitiveness of practical teaching are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of sensor calibration, and in particular to a multi-sensor external parameter calibration system and method based on a robotic arm and an adjustable slide rail. Background Art

[0002] During the extrinsic parameter calibration process, ensuring precise alignment between sensors and between sensors and the vehicle coordinate system is crucial for application scenarios such as autonomous driving and training and accuracy verification of robotic perception systems.

[0003] Accurate extrinsic parameter calibration improves the accuracy of multi-sensor data fusion, enhances environmental perception, and ensures the system makes reliable decisions in complex scenarios. It directly impacts the effectiveness of key functions such as object detection and path planning, and is fundamental to ensuring autonomous driving safety and the success rate of robotic missions. Through meticulous calibration, the robustness and stability of the system can be significantly improved, enabling more precise operation and control.

[0004] Traditional multi-sensor calibration relies on calibration plates or manual measurements, and its calibration goal is to "calculate" an external parameter. The calibration results of this method cannot directly verify the true value, and the source of the error is difficult to trace: In existing robotic arm calibration schemes, the relative positions of sensors are fixed, and it is impossible to implement a benchmark true value setting and comparison mechanism. Specifically, traditional methods, such as chessboard-based calibration, can solve the relative posture relationship between sensors, but this calculation result itself is the final answer, lacking an independent, physical "true value" for comparison and verification. Therefore, when an error occurs in the calibration, it is impossible to determine whether it is a problem with the algorithm itself, an error in sensor observation, or an error in the robotic arm movement, and the source of the error is difficult to trace. In addition, manually adjusting the sensor position is inefficient and cannot guarantee repeatability accuracy. Summary of the Invention

[0005] The present invention aims to provide a multi-sensor extrinsic parameter calibration system and method based on a robotic arm and an adjustable slide rail. By cleverly setting the robotic arm and the slide rail with a scale, the extrinsic parameter reference value (extrinsic parameter 0) can be preset. The robotic arm drives the multi-degree-of-freedom movement of the slide rail and the sensor, automatically acquiring multi-view observation data and fusion analysis, achieving the verifiability of the calibration results and the intuitiveness of practical training, while ensuring the repeatability and accuracy of the calibration.

[0006] The basic solution provided by the present invention is: a multi-sensor external parameter calibration system based on a robotic arm and an adjustable slide rail, comprising: A calibration target unit, comprising a calibration stand and a target fixed on the calibration stand and provided with a target mark; The robotic arm slide rail unit includes an electrically connected control system and a robotic arm, with a scaled slide rail at the end of the robotic arm. The control system controls the robotic arm to drive the slide rail to multiple positions and obtains corresponding position data. The sensor unit includes at least two sensors connected by a slide rail. The relative positions of the sensors are adjusted along the slide rail and locked at a preset scale, and the target object observation data corresponding to each posture is collected by the sensors. The data processing unit includes a data input module and a data calculation module. The data input module is used to exchange data with the calibration target unit, the robotic arm slide unit and the sensor unit. The data calculation module is used to solve the corresponding extrinsic parameter estimation value based on each posture data and its corresponding target object observation data, combined with the constructed calibration platform coordinate system, and optimize to obtain the final extrinsic parameter value, compare the final extrinsic parameter value with the preset extrinsic parameter reference value determined according to the preset scale to calculate the error.

[0007] The present invention also provides a multi-sensor external parameter calibration method based on a robotic arm and an adjustable slide rail, using a multi-sensor external parameter calibration system based on a robotic arm and an adjustable slide rail, the calibration method includes: S1, adjust the relative position of the sensor along the slide rail and lock it at a preset scale to determine the preset external parameter reference value; establish a calibration gantry coordinate system and determine the coordinates of the target object and the target object mark in the calibration gantry coordinate system; S2, using the control system to control the robotic arm to drive the slide to multiple positions and obtain the corresponding position data; the sensor collects the target object observation data corresponding to each position; using each position data and its corresponding target object observation data, combined with the calibration gantry coordinate system, the extrinsic parameter estimation value corresponding to each position is solved; S3, optimize the extrinsic parameter estimation values ​​corresponding to all postures and obtain the final extrinsic parameter value, and compare the final extrinsic parameter value with the preset extrinsic parameter reference value to calculate the error.

[0008] The working principle and advantages of the present invention are: The calibration goal in traditional general calibration scenarios is to "calculate" an external parameter without the need for verification. In the conventional application of this general calibration scenario, the traditional calibration method lacks a calibration idea of ​​verifying the closed loop.

[0009] The present invention pays special attention to two special application scenarios: automated calibration and practical training. In the automated calibration production line, it is necessary to ensure that the accuracy of the calibration system itself is reliable and verifiable; in practical training, students need to intuitively understand the physical meaning of external parameters and be able to verify the effectiveness of the calibration algorithm by themselves. It is the needs of these two special scenarios that have given rise to the idea that a physical true value must be introduced for comparison, which is not a priority in general calibration scenarios. Therefore, the goal of the present invention is not only to "calculate", but also to "verify" the accuracy of the calculation result, forming a closed loop of "setting true value → measurement → calculation → verification".

[0010] This invention addresses the technical pain points of existing technologies, including the inability to self-verify and trace errors, by cleverly designing a graduated slide rail as a physical truth carrier. Compared to existing technologies, this invention's graduated slide rail allows for presetting of an external reference value (external reference 0), providing a verifiable truth benchmark for calibration results. Combined with a high-precision robotic arm driving the slide rail, it further drives the multi-view motion of the sensor and the known position of the calibration target, enabling the collection, automated fusion optimization, and error analysis of multi-view target observation data. This results in closed-loop verification of calibration results and traceability of the calibration process.

[0011] Furthermore, the scale design on the slide rail not only allows for a preset external parameter baseline value (external parameter 0), but also allows trainees to manually change the sensor position (changing external parameter 0) to simulate disturbance scenarios. The resulting error results visually demonstrate the calibration algorithm's adaptability to dynamic external parameters. Furthermore, the introduction of the scale allows for preset sensor adjustments, ensuring calibration repeatability and accuracy.

[0012] Through the present invention, the calibration accuracy can reach translation error ≤ 1mm and rotation error ≤ 0.5°. It also supports automated calibration and practical training, with efficiency improved by more than 70% compared with traditional methods, and has a wide range of applicable scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A schematic structural diagram of a multi-sensor extrinsic parameter calibration system based on a robotic arm and an adjustable slide rail provided in an embodiment of the present invention; Figure 2 A schematic flow chart of a multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail provided in an embodiment of the present invention.

[0014] The symbols in the drawings of the specification include: calibration platform 1, target object 2, robotic arm 3, robotic arm base 31, slide rail 4, laser radar 5, camera 6, and data processing unit 7. DETAILED DESCRIPTION

[0015] The following is a further detailed description through specific implementation methods: The embodiment is basically as shown in the attached Figure 1 Shown: A multi-sensor external parameter calibration system based on a robotic arm and adjustable slide rails, including: The calibration target unit includes a calibration stand 1 and a target 2 fixed on the calibration stand 1 and provided with a target mark; The robot arm slide rail unit includes an electrically connected control system and a robot arm 3, and a slide rail 4 with a scale is provided at the end of the robot arm 3; the control system controls the robot arm 3 to drive the slide rail 4 to multiple positions and obtains corresponding position data; The sensor unit includes at least two sensors connected by a slide rail 4. By adjusting the relative positions of the sensors along the slide rail 4 and locking them at a preset scale, the sensors collect observation data of the target object corresponding to each posture; The data processing unit 7 includes a data input module and a data calculation module. The data input module is used to interact with the calibration target unit, the robot slide rail unit and the sensor unit ( Figure 1 The mid-bend curve is only used as a diagram for data interaction. The data calculation module is used to solve the corresponding extrinsic parameter estimation value based on each pose data and its corresponding target object observation data, combined with the constructed calibration gantry coordinate system, and optimize to obtain the final extrinsic parameter value. The final extrinsic parameter value is compared with the preset extrinsic parameter reference value determined according to the preset scale to calculate the error.

[0016] like Figure 2 As shown, this embodiment also provides a multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail. Using a multi-sensor extrinsic parameter calibration system based on a robotic arm and an adjustable slide rail, the calibration method includes: S1, adjust the relative position of the sensor along the slide rail 3 and lock it at a preset scale to determine the preset external parameter reference value; establish a calibration gantry coordinate system and determine the coordinates of the target object 2 and the target object mark in the calibration gantry coordinate system; S2, using a control system to control the robotic arm 3 to drive the slide rail 4 to multiple positions and obtain corresponding position data; the sensor collects the target object observation data corresponding to each position; using each position data and its corresponding target object observation data, combined with the calibration platform coordinate system, solve the extrinsic parameter estimation value corresponding to each position; S3, optimize the extrinsic parameter estimation values ​​corresponding to all postures and obtain the final extrinsic parameter value, and compare the final extrinsic parameter value with the preset extrinsic parameter reference value to calculate the error.

[0017] Specifically, an embodiment is given below for application illustration: Hardware preparation: A six-axis robotic arm (repeat positioning accuracy ≤ 0.05 mm) is selected, and the robotic arm 3 is placed on the robotic arm base 31 .

[0018] The slide rail 4 is a linear rail with a length of 800-1000 mm and a scale accuracy of 0.1 mm. All sensors are slidably and lockably connected to the slide rail via pre-set connectors. The pre-set connectors or fixtures can adopt existing structures that can achieve the sliding and locking effects of the present invention and are not limited to specific structures.

[0019] Two sensors are set up, namely a laser radar 5 and a camera 6, among which the laser radar 5 adopts the 16-line model product of RoboSense, and the camera 6 adopts the Intel RealSense D455 model product; the sensor data output end is connected to the data input module, and the access method is realized by adopting existing technology.

[0020] Target object 2 is a geometric feature object (such as a cube, calibration plate). Target object 2 is a cube with a side length of 30 cm and an AprilTag mark on the surface. The coordinates of its center point are calibrated by a three-dimensional coordinate measuring machine. .

[0021] The host computer is used to integrate and execute the data processing unit functions.

[0022] Calibration process: In S1, the external parameter reference value (external parameter 0) is preset and the calibration gantry coordinate system is established; 1) Move the lidar 5 and camera 6 along the slide rail 4 to the scale mark position and record their relative position (external reference 0).

[0023] Specifically, the operator adjusts the position of the LiDAR 5 (sensor A) and the camera 6 (sensor B) along the slide rail 4 and locks them on the preset connector of the slide rail 4. For example, the camera 6 is fixed at the "0.0mm" scale of the slide rail 4, and the LiDAR 5 is fixed at the "500.0mm" scale of the slide rail 4. During installation, there are clear requirements for the orientation of the sensors: the coordinate systems of the two sensors are pre-set to be aligned with the coordinate system of the slide rail. Specifically, assuming that the direction of the slide rail 4 (which can be understood as the sliding extension direction of the slide rail 4) is the X-axis, and the vertical plane of the slide rail 4 is the Z-axis, then the Z-axes of the optical centers (or scanning centers) of the camera 6 and the LiDAR 5 are required to be parallel to the Z-axis of the slide rail 4, and the Y-axes are also parallel to each other.

[0024] At this point, the operator manually inputs the scale position data (i.e., scale readings) of the two sensors through the user interface of the host computer software (i.e., data input module). The internal program of the data calculation module automatically constructs the preset external parameter reference transformation matrix based on the scale position data of the sensors and the coordinate system definition of the slide 4 itself (for example, the direction of the slide 4 is the X axis). . This matrix usually includes a translation vector (such as (unit: meters) and a unit rotation matrix representing the direction alignment .this It is stored in the data processing unit as the "true value" (i.e. the preset external parameter reference value) for the final error calculation.

[0025] 2) Measure the physical distance and other relevant data between the target object 2 and the robot arm base 31 using a ruler or a laser rangefinder (arrange the measurement data according to the requirements of characterizing the target object and the target object mark position). The operator manually inputs the physical distance and other relevant data through the user interface of the host computer software (i.e., the data input module). The internal program of the data calculation module calculates the coordinates of the target object and the target object mark in the calibration rig coordinate system based on the relevant data characterizing the target object and the target object mark position and the definition of the calibration rig's own coordinate system.

[0026] In S2, the control system is used to control the robot arm 3 to drive the slide rail 4 to multiple positions, and the following operations are performed at each position: S21, robot arm 3 outputs end pose data:

[0027] S22, two sensors (lidar 5 and camera 6) respectively collect observation data of the target object under the perspective of the posture; Specifically, the laser radar 5 collects the three-dimensional point cloud of the target object 2, and the camera 6 detects and recognizes the AprilTag mark on the target object 2 and obtains the two-dimensional pixel coordinates of each corner point thereof.

[0028] S23, calculating the position of the sensor relative to the output end of the manipulator based on the target object observation data at the pose viewing angle and the calibration gantry coordinate system; Specifically, the position of the camera 6 relative to the output end of the robot 3 under the condition of posture i is calculated by the perspective-n-point (PnP) algorithm. The PnP algorithm is a classic method for solving the camera pose estimation problem. Specifically, the PnP algorithm aims to calculate the camera's position and orientation relative to known three-dimensional spatial points and their projections on the two-dimensional image plane. This involves determining the camera's extrinsic parameters (including the rotation matrix and translation vector). This technique has widespread applications in computer vision, robotics, augmented reality (AR), autonomous driving, and other fields.

[0029] The position of the LiDAR 5 relative to the output end of the manipulator under the condition of posture i is calculated by the Iterative Closest Point (ICP) algorithm. The ICP algorithm is a widely used technique in the field of calibration, especially when it comes to the registration and alignment of 3D point cloud data. The ICP algorithm is primarily used to accurately align two 3D point cloud datasets. By minimizing the distance between corresponding points, it finds the optimal spatial transformation (including rotation and translation) to ensure that one point cloud matches the other as closely as possible.

[0030] In this process, the "calibration gantry coordinate system" plays the role of the "world coordinate system". The specific data processing process is as follows: 1. Using the PnP algorithm: Given the 3D coordinates of each corner point on the AprilTag on the target 2 in the calibration gantry coordinate system, and the 2D pixel coordinates of these corner points in the camera 6 image, the PnP algorithm uses these two sets of correspondences to directly calculate the position and posture of the camera 6 in the calibration gantry coordinate system. .

[0031] 2. Using the ICP algorithm: The LiDAR 5 collects a 3D point cloud of the target. The coordinates of the complete 3D model of the target 2 in the calibration rig coordinate system are known. The ICP algorithm iteratively matches the collected point cloud with the known model to calculate the position and posture of the LiDAR 5 in the calibration rig coordinate system. .

[0032] 3. Calculate relative pose: Due to the pose of robot arm 3 It is relative to the coordinate system of the robot base 31, and the conversion relationship between the base coordinate system and the calibration gantry coordinate system (world coordinate system) is usually It is pre-calibrated and fixed. Therefore, the position of the sensor relative to the end of the robot arm 3 can be calculated, for example:

[0033] Where, represents the position data of the camera 6 relative to the end of the robotic arm 3 under the condition of posture i; Indicates that the robot arm 3 outputs the end position i data; Indicates the conversion relationship between the robot arm base coordinate system and the calibration gantry coordinate system; Represents the position and posture of camera 6 in the calibration gantry coordinate system.

[0034]

[0035] Where, represents the position of the laser radar 5 relative to the output end of the manipulator 3 under the condition of posture i; Represents the position and posture of the laser radar 5 in the calibration gantry coordinate system.

[0036] S24, calculate the estimated value of the external parameter between the laser radar 5 and the camera 6 by the following formula :

[0037] Where, represents the estimated value of the external parameter under the condition of pose i (indicating the transformation from the lidar coordinate system to the camera coordinate system); represents the position data of the camera 6 relative to the end of the robotic arm 3 under the condition of posture i; Represents the position data of the laser radar 5 relative to the end of the robot arm 3 under the condition of posture i.

[0038] In S3, the results are verified; S31, estimation of multiple groups of external parameters Perform optimization (such as weighted least squares method) to obtain the final external parameter value .

[0039] Specifically, the Geodesic Mean algorithm can be used to fuse multiple groups , get the final external parameter value Lie group averaging provides an effective method for processing data in non-Euclidean spaces. It takes into account the geometric structure of these special spaces to ensure that the results remain within the space and can reflect the intrinsic relationships between data points, ensuring that the results are both physically meaningful and have good mathematical properties. This is crucial for many applications involving complex transformations and high-dimensional data.

[0040] For a set of data points ( {g_1, g_2, ..., g_n} ) belonging to a Lie group ( G ), its Lie group mean can be defined as:

[0041] Here, d(·,·) represents the geodesic distance function on the Lie group. In actual calculations, numerical optimization techniques are often needed to approximate the solution to this minimization problem.

[0042] S32, will Compare with the preset extrinsic parameter reference value (extrinsic parameter 0), calculate the error (such as the Euclidean distance of the translation vector and the angular axis error of the rotation matrix) to verify the accuracy of the calibration algorithm.

[0043] Specifically, the translation error (in mm) and rotation error (in degrees) relative to the external parameter 0 can be calculated using the following formula:

[0044]

[0045] Where, The translation vector representing the final extrinsic parameter value; The rotation matrix representing the final extrinsic parameter value; A translation vector representing a preset extrinsic reference value; A rotation matrix representing the preset extrinsic reference value.

[0046] In this embodiment, It represents the final optimized external parameter transformation matrix, which describes the transformation relationship from the laser radar coordinate system to the camera coordinate system. It consists of the rotation part and pan part composition. for The translation vector (3x1 can be used). for The rotation matrix (3x3 can be used).

[0047] Indicates the preset external parameter reference value (i.e., "true value") set by the scale of slide rail 4. yes The translation vector of . yes The rotation matrix of . yes The transpose matrix of a rotation matrix is ​​equal to its inverse.

[0048] The following calibration test illustrates the entire process from setting the true value, single measurement, fusion optimization to final error verification.

[0049] S1: As mentioned above, the camera 6 is set at the scale of 0mm and the laser radar 5 is set at the scale of 500mm, and the two are aligned. The data processing unit generates and stores the preset external parameter reference value (external parameter 0):

[0050] (Identity Matrix) S2 (taking the i-th posture as an example): The robot arm 3 moves to posture i and outputs the posture of its end relative to the base .

[0051] After the sensor collects data, the data processing unit calculates the position of the camera 6 and the lidar 5 relative to the end of the robotic arm 3 through the PnP and ICP algorithms. and .

[0052] According to the formula , calculate the estimated value of the external parameters under the current posture. For example, the calculation result is: ,in .

[0053] S31 (Optimization): Repeat step S2 5 times to obtain 5 sets of external parameter estimates , the Lie group average algorithm is used to fuse and optimize these five sets of data to obtain the final external parameter value . Assume that the optimized result is: (rice) is a rotation matrix that is very close to the identity matrix.

[0054] S32 (calculation error): Substitute the final external parameter value and the preset external parameter reference value into the error formula: Translation error:

[0055] Rotation error: First calculate .because is a unit matrix, so ;

[0056] Assumptions The calculated result is 2.99985, so the rotation error is about 0.3 degrees.

[0057] The translation error and rotation error results obtained from the five groups of experiments are shown in Table 1: Table 1 Calibration test error results

[0058] The results of five independent repeated tests in Table 1 show that the translational errors calibrated by this system and method are all less than 1.0 mm, and the rotational errors are all less than 0.5°. This demonstrates the high calibration accuracy and repeatability of the present invention. The calculated results are highly consistent with the physically established benchmark values, thus verifying the effectiveness and reliability of the proposed system and method.

[0059] This system also demonstrates its unique intuitive advantages in practical training. For example, the following practical training tasks can be designed: 1. Benchmark calibration: The trainees first fixed the lidar 5 and camera 6 at the 0mm and 500mm scales of the slide rail according to the standard process, performed an automated calibration, and obtained the reference error shown in Experimental Group 1 (0.8mm translation, 0.3° rotation).

[0060] 2. Simulate disturbance: The instructor asks the trainee to "simulate an installation deviation". The trainee manually moves the laser radar 5 along the slide rail 4 by +20mm and locks it at the 520mm scale. At this time, the new external reference value The x-component becomes 0.52 meters. 3. Perturbation recalibration and verification: The trainee runs the automated calibration procedure again. The final translation vector calculated by the system The x-component of the coordinate system will be very close to 0.52 meters. At the same time, by comparing the new calculation results with the new baseline values, the reported translation and rotation errors will remain at a small level of around 1mm and 0.5°. Through this process, students were able to personally operate and intuitively see that the system accurately measured the 20mm physical disturbance they introduced, and that the intrinsic accuracy of this calibration method remained stable regardless of changes in the relative position of the sensors. This greatly deepened their understanding of the physical significance of "external parameters" and their confidence in the robustness of the calibration algorithm.

[0061] The multi-sensor external parameter calibration system and method based on a robotic arm and an adjustable slide provided in this embodiment can pre-set an external parameter reference value (external parameter 0) by designing a scaled slide, providing a verifiable true value reference for the calibration results. The high-precision robotic arm drives the slide to further drive the multi-perspective movement of the sensor and the known position of the calibration target, thereby enabling the collection, automated fusion optimization, and error analysis of multi-perspective target observation data, achieving closed-loop verification of the calibration results and traceability of the calibration process. Furthermore, the scale design of the slide not only enables the preset external parameter reference value (external parameter 0), but also enables trainees to manually change the sensor position (changing external parameter 0) to simulate disturbance scenarios. The resulting error results can intuitively demonstrate the adaptability of the calibration algorithm to dynamic external parameters, and are applicable to a wide range of scenarios.

[0062] The above is only an embodiment of the present invention. Common knowledge such as the specific structure and characteristics of the scheme is not described in detail here. Ordinary technicians in the relevant field are aware of all common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the guidance of this application. Some typical well-known structures or well-known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A multi-sensor external parameter calibration system based on a robotic arm and an adjustable slide rail, characterized by: include: A calibration target unit, comprising a calibration stand and a target fixed on the calibration stand and provided with a target mark; The robotic arm slide rail unit includes an electrically connected control system and a robotic arm, with a scaled slide rail at the end of the robotic arm. The control system controls the robotic arm to drive the slide rail to multiple positions and obtains corresponding position data. The sensor unit includes at least two sensors connected by a slide rail. The relative positions of the sensors are adjusted along the slide rail and locked at a preset scale, and the target object observation data corresponding to each posture is collected by the sensors. The data processing unit includes a data input module and a data calculation module. The data input module is used to exchange data with the calibration target unit, the robotic arm slide unit and the sensor unit. The data calculation module is used to solve the corresponding extrinsic parameter estimation value based on each posture data and its corresponding target object observation data, combined with the constructed calibration platform coordinate system, and optimize to obtain the final extrinsic parameter value, compare the final extrinsic parameter value with the preset extrinsic parameter reference value determined according to the preset scale to calculate the error.

2. The multi-sensor external parameter calibration system based on a robotic arm and an adjustable slide rail according to claim 1, characterized in that: The data input module is used to input the scale position data of the sensor; the data calculation module is used to automatically construct a preset external parameter reference transformation matrix representing the preset external parameter reference value based on the scale position data of the sensor and the coordinate system definition of the slide rail itself, including a translation vector and a unit rotation matrix representing direction alignment.

3. The multi-sensor external parameter calibration system based on a robotic arm and an adjustable slide rail according to claim 1, characterized in that: The data input module is used to input relevant data representing the position of the target object and the target mark; the data calculation module is used to calculate the coordinates of the target object and the target mark in the calibration rig coordinate system based on the relevant data representing the position of the target object and the target mark and the definition of the calibration rig's own coordinate system.

4. The multi-sensor external parameter calibration system based on a robotic arm and an adjustable slide rail according to claim 1, characterized in that: The slide rail is a linear slide rail, and all sensors are slidably and lockably connected to the slide rail through preset connectors.

5. A multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail, characterized in that: Utilizing the multi-sensor extrinsic parameter calibration system based on a robotic arm and an adjustable slide rail according to any one of claims 1 to 4, the calibration method includes: S1, adjust the relative position of the sensor along the slide rail and lock it at a preset scale to determine the preset external parameter reference value; establish a calibration gantry coordinate system and determine the coordinates of the target object and the target object mark in the calibration gantry coordinate system; S2, using the control system to control the robotic arm to drive the slide to multiple positions and obtain the corresponding position data; the sensor collects the target object observation data corresponding to each position; using each position data and its corresponding target object observation data, combined with the calibration gantry coordinate system, the extrinsic parameter estimation value corresponding to each position is solved; S3, optimize the extrinsic parameter estimation values ​​corresponding to all postures and obtain the final extrinsic parameter value, and compare the final extrinsic parameter value with the preset extrinsic parameter reference value to calculate the error.

6. The multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail according to claim 5, characterized in that: When two sensors are set in S1, namely a lidar and a camera, in S2, the estimated value of the external parameter is calculated by the following formula: Where, represents the estimated value of the external parameters under the condition of posture i; Represents the position data of the camera relative to the end of the robotic arm under the condition of posture i; Represents the position data of the lidar relative to the end of the robotic arm under the condition of posture i.

7. The multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail according to claim 6, characterized in that: In S2, the perspective-n-point algorithm is used to calculate the camera's position in the calibration gantry coordinate system based on the three-dimensional coordinates of each corner point of the target object in the calibration gantry coordinate system and the corresponding relationship between the two-dimensional pixel coordinates of these corner points in the camera image. ;Then : Where, Represents the position data of the camera relative to the end of the robotic arm under the condition of posture i; Indicates the output end position i data of the robot arm; Indicates the conversion relationship between the robot arm base coordinate system and the calibration gantry coordinate system; Indicates the position and orientation of the camera in the calibration gantry coordinate system.

8. The multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail according to claim 6, characterized in that: In S2, the iterative closest point algorithm is used to iteratively match the three-dimensional point cloud of the target object collected by the lidar and the coordinates of the three-dimensional model of the target object in the calibration rig coordinate system, and the position and posture of the lidar in the calibration rig coordinate system are calculated. ; Then use the following formula to calculate : Where, represents the position of the lidar relative to the output end of the manipulator under the condition of posture i; Indicates the output end position i data of the robot arm; Indicates the conversion relationship between the robot arm base coordinate system and the calibration gantry coordinate system; Represents the position and posture of the lidar in the calibration rig coordinate system.

9. The multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail according to claim 5, characterized in that: In S3, the Lie group averaging algorithm is used to fuse and optimize the extrinsic parameter estimates corresponding to all postures.

10. The multi-sensor extrinsic parameter calibration method based on a robotic arm and an adjustable slide rail according to claim 5, characterized in that: In S3, the translation error and rotation error are calculated by the following formula: Where, The translation vector representing the final extrinsic parameter value; The rotation matrix representing the final extrinsic parameter value; A translation vector representing a preset extrinsic reference value; A rotation matrix representing the preset extrinsic reference value.

Images