A large-scale three-dimensional tracking method based on multi-camera fusion

Through the multi-camera fusion system, the problem of cumbersome pasting reference points and low measurement efficiency in measurement of large industrial parts is solved, and high-precision, large-scale, three-dimensional measurement is achieved, which simplifies the process and improves portability.

CN114494362BActive Publication Date: 2024-11-22WUHAN POWER3D TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210049247.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-17
Publication Date
2024-11-22
Estimated Expiration
2042-01-17

AI Technical Summary

Technical Problem

The existing large-scale three-dimensional measurement technology requires the pasting of a large number of reference points when measuring large industrial parts. The process is cumbersome and the measurement efficiency is inefficient, and it is not suitable for all workpieces, especially for insufficient portability and accuracy.

Method used

Using a large-scale three-dimensional tracking method based on multi-camera fusion, including servo motors, movable cameras, search cameras, rotary tables and reference calibration plates, the rotation angles of rotary tables and servo motors are calculated through the multi-camera system tracking and calibration set of mark points. , ensure that the set of marking points is always within the field of view of the movable camera, and record the movable camera position changes by referring to the camera to synthesize the measurement results.

Benefits of technology

It realizes high-precision, large-scale and three-dimensional measurements, simplifies the measurement process, improves measurement efficiency, and has good portability in the system. It is suitable for all kinds of large parts, ensuring the combination of high precision and large field of view.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114494362B_ABST
    Figure CN114494362B_ABST
Patent Text Reader

Abstract

The present invention discloses a large-scale three-dimensional tracking method based on multi-camera fusion, the method comprising: building a large-scale three-dimensional tracking system based on multi-camera fusion; calibrating the initial position and posture of a movable camera by a reference camera; tracking a marker point set by searching the camera, and placing the marker point set in the field of view of the movable camera by rotating a rotating table and a servo motor; continuously measuring the position and posture of the marker point set at different measuring points by the movable camera, and at the same time, the reference camera obtains the position and posture change information of the movable camera through a reference calibration plate fixedly connected to the movable camera; and splicing the measurement results of the movable camera into the initial coordinate system of the movable camera according to the position and posture information of the marker point set at different measuring points measured by the movable camera and the position and posture change information of the movable camera. The present invention adopts multiple monocular cameras for target tracking measurement, and adopts a mechanical rotating table to expand the camera field of view, and can be applied to high-precision three-dimensional measurement of various large parts.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of three-dimensional measurement, and in particular relates to a large-scale three-dimensional tracking method based on multi-camera fusion. Background Art

[0002] At present, in the field of industrial parts measurement, visual measurement has the advantages of fast measurement speed, large data volume, and non-contact, and has been widely used. However, for large industrial parts, visual measurement is limited by factors such as depth of field and field of view. It is often necessary to paste marking points on the surface of the workpiece and then combine multiple measurement results. However, the process of pasting marking points is cumbersome, and the density of marking points needs to be reasonably arranged. The measurement efficiency is low, and it does not meet the automated measurement requirements of current industrial production. At the same time, it is not suitable for the measurement of workpieces that are not suitable for pasting marking points.

[0003] To solve this problem, the existing solutions are mainly divided into two types: three-dimensional measurement based on precision machinery and measurement based on rear tracking. Among them, precision machinery drives the measurement equipment to move, and real-time feedback of the measurement equipment posture is provided, and finally the measurement results are spliced ​​together. However, the precision machinery is expensive and not portable enough; rear tracking is to rigidly fix the marker point structure and the measurement system, and the rear tracking system measures the changes in the marker point structure posture, and then splices the measurement results. This method has high accuracy and good portability, and has been widely used.

[0004] However, due to the limited field of view of the rear tracking system, when measuring large workpieces, it is necessary to paste auxiliary markers and adjust the position of the tracking system. The measurement steps are complicated and the accuracy is reduced. Therefore, a measurement system with high accuracy, large measurement range and good portability is needed. Summary of the invention

[0005] In view of this, the present invention proposes a large-scale three-dimensional tracking method based on multi-camera fusion, which is used to solve the problems of the existing large-scale three-dimensional measurement that requires pasting a large number of reference points and the complicated measurement process, and provides an effective solution for large-size three-dimensional measurement.

[0006] The present invention discloses a large-scale three-dimensional tracking method based on multi-camera fusion, the method comprising:

[0007] A large-scale three-dimensional tracking system based on multi-camera fusion is constructed, the system comprising a servo motor, a movable camera, a search camera, a reference camera, a rotating stage, a reference calibration plate and a base; wherein the rotating stage is mounted on the base, the reference camera is fixedly connected to the base, the servo motor and the search camera are fixed on the rotating stage, the movable camera is fixedly connected to the rotating shaft of the servo motor, and the reference calibration plate is fixed under the movable camera;

[0008] Calibrate the initial pose of the movable camera through the reference camera;

[0009] The search camera tracks the landmark point set, calculates the rotation angle of the rotating stage and the servo motor, and the landmark point set is always placed within the field of view of the movable camera through the rotation of the rotating stage and the servo motor;

[0010] The movable camera continuously measures the pose of the marker point set at different measuring points, and converts the pose information of the marker point set into the movable camera coordinate system. Meanwhile, the reference camera obtains the pose change information of the movable camera through the reference calibration plate under the movable camera.

[0011] According to the pose information of the landmark point set of different measuring points measured by the movable camera and the pose change information of the movable camera, the measurement result of the movable camera is spliced ​​into the initial pose of the movable camera.

[0012] On the basis of the above technical solution, preferably, the calibrating the initial position and posture of the movable camera by using the reference camera specifically includes:

[0013] The reference camera calibrates the initial position and posture of the movable camera by taking pictures of a reference calibration plate fixedly connected to the movable camera.

[0014] On the basis of the above technical solution, preferably, before tracking the marker point set by the search camera, the method further includes calibrating the relative position of the search camera and the movable camera, and calibrating the rotation axis of the rotating table and the servo motor:

[0015] The search camera and the movable camera are calibrated through an external calibration board to obtain the posture conversion relationship between the two:

[0016]

[0017] in, To search for the camera relative to the movable camera pose matrix, To search for the pose matrix of the camera relative to the reference calibration plate, is the pose matrix of the reference calibration plate relative to the movable camera; under this condition, the rotation angles of the rotating table and the servo motor are α 0 , β 0 ;

[0018] When calibrating the rotating axis of the rotating stage and servo motor, the movable camera or search camera continuously takes pictures of an external calibration plate during the rotation of the rotating stage or servo motor, and obtains the l positions of the optical center of the movable camera or search camera in the coordinate system of the external calibration plate, and then calculates the rotating axis and a point on the rotating axis:

[0019] n=(u,v,w)

[0020] o=(a,b,c)

[0021] Where n is the unit rotation axis vector, and o is a point on the rotation axis;

[0022] According to the Rodriguez formula, the transformation formula of the rotation matrix R after the movable camera or the search camera rotates around the axis by an angle θ is obtained:

[0023] R=cosθI+(1-cosθ)nn T +sinθn ∧ Where θ is the rotation angle, I is the identity matrix, n ∧ represents the antisymmetric matrix of n; according to the conversion formula, the rotation matrix R of the movable camera or the search camera rotating around the axis by an angle θ is obtained θ and the translation matrix t θ .

[0024] On the basis of the above technical solution, preferably, the tracking of the marker point set by the search camera, calculating the rotation angle of the rotating table and the servo motor, and placing the marker point set always within the field of view of the movable camera by rotating the rotating table and the servo motor specifically includes:

[0025] The rotating stage drives the search camera to rotate until the set of landmark points appears in the field of view of the search camera;

[0026] Calculate the centroid of the marker point set and obtain the coordinates of the centroid of the marker point set in the search camera coordinate system, recorded as (x, y, z);

[0027] According to the calibrated rotation axis of the rotating stage and servo motor and the position conversion relationship between the search camera and the movable camera, the coordinates (x′, y′, z′) of the center of gravity of the marker point in the initial position coordinate system of the movable camera are obtained as follows:

[0028]

[0029] Where α is the relative rotation of the rotating stage 0 The rotation angle of The angle between the rotary table and the servo motor is α 0 , β 0 On this basis, after the rotating platform rotates around the axis α, the transformation matrix of the movable camera relative to the initial position is R α is the corresponding rotation matrix, t α is the corresponding translation matrix; To search for the pose matrix of the camera relative to the reference calibration plate, is the pose matrix of the reference calibration plate relative to the movable camera;

[0030] The rotation angles of the rotating table and the servo motor are calculated respectively, and the rotation of the rotating table and the servo motor is controlled according to the rotation angles so that the center of gravity of the marker point set is directly opposite to the optical axis of the movable camera.

[0031] On the basis of the above technical solution, preferably, the calculating the rotation angles of the rotating table and the servo motor respectively, and controlling the rotation of the rotating table and the servo motor according to the rotation angles so that the center of gravity of the marker point set faces the optical axis of the movable camera specifically includes:

[0032] Assuming that the rotation angles of the rotating table and servo motor are α′ and β′ respectively, we can get:

[0033]

[0034] in, mα′ and tα′ are the rotation matrix and translation matrix corresponding to the rotation angle of the turntable of α′, Rβ′ and tβ′ are the rotation matrix and translation matrix corresponding to the rotation angle of the servo motor of β′, respectively. The above formula is simplified into a set of two-variable linear equations about α′ and β′:

[0035]

[0036] Among them, S 1 —S 8 As known constants, solve this set of two-variable linear equations to obtain the values ​​of α′ and β′;

[0037] The rotating table and the servo motor are controlled to rotate according to the values ​​of the rotation angles α′ and β′ so that the center of gravity of the marker point set is directly opposite to the optical axis of the movable camera.

[0038] On the basis of the above technical solution, preferably, according to the pose information of the marker point set of different measuring points measured by the movable camera and the pose change information of the movable camera, the measurement result of the movable camera is stitched into the initial pose of the movable camera, which specifically includes:

[0039] Suppose that the m poses of the landmark point set of different measuring points are obtained by continuous measurement by a movable camera, i=1,2,…,m, m is the total number of measurement points in the process of moving the marker point set. The reference camera measures the reference calibration plate fixed to the movable camera and obtains the pose change of the movable camera relative to the initial pose as ΔT i , then the pose information is stitched to the initial pose:

[0040]

[0041] The three-dimensional measurement results of the movable camera in the initial coordinate system are obtained.

[0042] On the basis of the above technical solution, preferably, the movable camera, the search camera and the reference camera are all monocular cameras.

[0043] Based on the above technical solution, preferably, the set of marker points is a light pen or other measuring equipment that carries three-dimensional information.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1) The present invention uses multiple monocular cameras for target tracking measurement, which can be applied to the three-dimensional measurement of various large parts. On the basis of ensuring accuracy, the size of the entire system is greatly reduced, and the portability is good.

[0046] 2) The present invention uses a monocular camera with a large field of view during measurement, and uses a mechanical rotating platform to expand the field of view of the camera, which can cover all ranges in front of the measuring device. At the same time, a combination of high precision and large field of view is adopted, that is, a high-precision movable camera is responsible for locating the marker point set, and a search camera with a large field of view searches for the marker point set through a mechanical platform, and feeds back the position of the marker point set to the high-precision camera for measurement. This method simplifies the measurement process, improves measurement efficiency, and is user-friendly.

[0047] 3) In the process of transforming the position and posture of the marker point set, the present invention adopts a dual rear positioning method, that is, there is a second tracking camera (reference camera), and the reference calibration plate pasted on the first tracking camera (movable camera) records the position and posture changes of the first tracking camera, and finally splices the three-dimensional measurement information into the initial coordinate system of the first tracking camera. While expanding the measurement range, only one coordinate transformation is added, ensuring high accuracy even in the case of large-scale measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 It is a schematic diagram of the structure of a large-scale three-dimensional tracking system based on multi-camera fusion of the present invention;

[0050] Figure 2 A schematic diagram of the structure of a light pen provided by an embodiment of the present invention;

[0051] Figure 3 It is a schematic diagram of the flow of the large-scale three-dimensional tracking method based on multi-camera fusion of the present invention;

[0052] Figure 4 It is a schematic diagram of the calibration of the rotating shaft of the rotating table and the servo motor of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] See also Figure 1 ,This invention proposes a large-scale three-dimensional tracking system based on multi-camera fusion, Figure 1 A schematic diagram of the system structure.

[0055] The system comprises a servo motor, a movable camera, a search camera, a reference camera, a rotating platform and a base, wherein the rotating platform is mounted on the base, the reference camera is fixedly connected to the base, the servo motor and the search camera are fixed on the rotating platform; the movable camera is fixedly connected to the rotating shaft of the servo motor, and a reference calibration plate is fixedly arranged on the movable camera;

[0056] The search camera is used to track and search the marker point set, and the marker point set is placed in the field of view of the movable camera through the rotation of the rotating table and the servo motor. The marker point set can be a light pen or other measuring equipment carrying three-dimensional information, such as Figure 2 FIG. 4 is a schematic diagram of a light pen used in an embodiment of the present invention.

[0057] The movable camera is used to measure the pose of the landmark point set and convert the pose information of the landmark point set into the movable camera coordinate system.

[0058] The reference camera is used to measure the change of the posture of the movable camera through a reference calibration plate fixed to the movable camera, and to stitch the measurement result of the movable camera to the initial posture of the movable camera.

[0059] The present invention uses multiple monocular cameras for target tracking measurement, combined with the rotation of the rotating table, uses a search camera with a large measurement range to locate the light pen, uses a high-precision movable camera to measure the light pen posture, uses a fixed-position reference camera to measure the change of the movable camera posture, and finally splices the measurement results to the initial posture of the movable camera. It can be applied to the three-dimensional measurement of various large parts, greatly reduces the size of the entire system on the basis of ensuring accuracy, and has good portability.

[0060] Based on the above-mentioned large-scale 3D tracking system based on multi-camera fusion, the present invention proposes a large-scale 3D tracking method based on multi-camera fusion, please refer to Figure 3 , the method comprising:

[0061] S1. Perform system calibration and calibrate the initial pose of the movable camera using a reference camera.

[0062] The system calibration specifically includes the search camera, movable camera calibration, relative position calibration between the search camera and movable camera, rotation axis calibration of the rotating stage and servo motor, etc. It specifically includes the following steps:

[0063] S11. Calibrate the initial pose of the movable camera

[0064] The reference camera takes a picture of the reference calibration plate fixedly connected to the movable camera to calibrate the initial position and posture of the movable camera, and the coordinate system at the initial position and posture of the reference camera is used as the initial coordinate system for assembling the measurement results;

[0065] S12. Calibrate the relative position of the search camera and the movable camera

[0066] The search camera and the movable camera are calibrated through an external calibration board to obtain the posture relationship between the two:

[0067]

[0068] in, To search for the pose matrix of the camera relative to the reference calibration plate, is the pose matrix of the reference calibration plate relative to the movable camera. Under this condition, the rotation angles of the rotating table and the servo motor are α 0 , β 0 Among them, the rotation angle of the servo motor refers to the rotation angle of the servo motor shaft. Since the servo motor shaft is fixedly connected to the movable camera, the rotation angle of the servo motor is the rotation angle of the movable camera in the vertical plane, and the rotation angle of the rotating table is the rotation angle of the entire servo motor, movable camera, and search camera in the horizontal plane.

[0069] S13, calibrate the rotating table and the rotating axis of the servo motor

[0070] like Figure 4 As shown, prepare another calibration plate as an external calibration plate. The search camera or movable camera continuously takes pictures of the external calibration plate during the rotation of the rotating table or servo motor to obtain the l positions of the optical center of the search camera or movable camera in the coordinate system of the external calibration plate, and then find the rotation axis and a point on the rotation axis:

[0071] n=(u,v,w)

[0072] o=(a,b,c)

[0073] Where n is the unit rotation axis vector, and o is a point on the rotation axis;

[0074] According to the Rodriguez formula, the transformation formula of the rotation matrix R after the movable camera or the search camera rotates around the axis by an angle θ is obtained:

[0075] R=cosθI+(1-cosθ)nnT+sinθn^

[0076] Where θ is the rotation angle, I is the unit matrix, and n∧ represents the antisymmetric matrix of n; according to the conversion formula, the rotation matrix and translation matrix R of the movable camera or the search camera rotating around the axis by an angle of θ are obtained. θ ,t θ .

[0077] S2. Track the marker point set through the search camera, feed back the marker point set pose information to the movable camera, calculate the rotation angle of the turntable and the servo motor, and keep the marker point set within the field of view of the movable camera through the rotation of the turntable and the servo motor.

[0078] S21, the rotating stage drives the search camera to rotate at an angle α until the set of landmark points appears in the field of view of the search camera. 0 The rotation angle based on is α;

[0079] S22, calculating the centroid of the marker point set, and obtaining the coordinates of the centroid of the marker point set in the search camera coordinate system, recorded as (x, y, z);

[0080] S23, according to the calibrated rotating table and the rotating axis of the servo motor and the posture conversion relationship between the search camera and the movable camera, the coordinates (x′, y′, z′) of the center of gravity of the marker point in the initial posture coordinate system of the movable camera are obtained as follows:

[0081]

[0082] in, The angle between the rotary table and the servo motor is α 0 , β 0 Based on the rotation table, after the rotation of the rotating table around the axis α, the transformation matrix of the movable camera relative to the initial position is R α is the corresponding rotation matrix, t α is the corresponding translation matrix, R α ,t α All are derived according to the Rodriguez formula in step S13; To search for the pose matrix of the camera relative to the reference calibration plate, is the pose matrix of the reference calibration plate relative to the movable camera, α is the position matrix of the rotating stage relative to α 0 Angle of rotation;

[0083] S24, respectively calculating the rotation angles of the rotating table and the servo motor, and controlling the rotation of the rotating table and the servo motor according to the rotation angles so that the center of gravity of the marker point set is directly opposite to the optical axis of the movable camera.

[0084] Specifically, assuming that the center of gravity of the marker point set is facing the optical axis of the movable camera, the rotation angles of the rotating table and the servo motor are α′ and β′ respectively, we can get:

[0085]

[0086] in, Rα′ and tα′ are the rotation matrix and translation matrix corresponding to the rotation angle of the rotating platform, Rβ′ and t β′ They are respectively the rotation matrix and the translation matrix corresponding to the rotation angle of the servo motor being β′, and are both obtained according to the Rodriguez formula in step S13.

[0087] The above formula can be simplified into a set of two-variable linear equations about α′ and β′:

[0088]

[0089] Among them, S 1 —S 8 As known constants, solving this set of two-variable trigonometric equations can yield the values ​​of α′ and β′;

[0090] The roots control the rotation angles α′ and β′ of the rotating table and the servo motor respectively, so that the center of gravity of the marker point set faces the optical axis of the movable camera.

[0091] S3. While continuously measuring the poses of the marker point sets at different measuring points through the movable camera, the reference camera obtains the pose change information of the movable camera through the reference calibration plate under the movable camera.

[0092] The PNP method is used to obtain the pose relationship between the marker point set and the movable camera. The marker point set moves to the next measuring point, and the movable camera continues to obtain the pose of the marker point set at different measuring points. At the same time, the reference camera uses the PNP method to obtain the pose relationship between the reference calibration plate and the reference camera, thereby obtaining the pose change information of the movable camera relative to the initial pose when measuring the marker point sets at different measuring points.

[0093] The present invention uses a monocular camera with a large field of view to track the marker point set during measurement, and uses a mechanical rotating platform to expand the camera's field of view, which can cover all ranges in front of the measuring device. At the same time, a combination of high precision and large field of view is adopted, that is, a high-precision movable camera is responsible for locating the marker point set, and a search camera with a large field of view searches for the marker point set through a mechanical platform, and the marker point set position information is fed back to the high-precision movable camera for measurement, which simplifies the measurement process and improves the measurement efficiency.

[0094] S4. According to the pose information of the landmark point set measured by the movable camera and the pose change information of the movable camera, the measurement result of the movable camera is stitched into the initial pose of the movable camera.

[0095] by Figure 2 Taking the light pen shown in the figure as an example, the set of landmark points is all the landmark points on the light pen. When the light pen is measuring, the movable camera identifies the landmark points on the light pen and obtains the conversion relationship between the light pen and the movable camera coordinate system, thereby converting the coordinates of the light pen tip to the camera coordinate system. Assume that the coordinates of the pen tip in the light pen coordinate system are (x 0 ,y 0 ,z 0 ), the position transformation relationship between the light pen coordinate system and the movable camera coordinate system at the measuring point i during measurement is: The pose change information of the movable camera relative to the initial pose is ΔT i , thus obtaining the coordinates of the pen tip in the movable camera coordinate system at the measuring point i:

[0096]

[0097] Assuming that the initial pose of the movable camera is pose 1, the m poses of the landmark point set are obtained by continuous measurement of the movable camera, which are: R i is the corresponding rotation matrix, t i is the corresponding translation matrix, i = 1, 2…, m, m is the total number of measurement points in the process of moving the marker point set, and the reference camera measures the reference calibration plate fixed to the movable camera to obtain the pose change information of the movable camera relative to the initial pose as ΔT i , i = 1, 2…, m, then the pose information of m measuring points is stitched together into pose 1:

[0098]

[0099] Finally, the three-dimensional measurement result of the movable camera in the initial coordinate system is obtained.

[0100] The present invention tracks the change of the position of the movable camera through the reference camera while tracking the search mark point set and positioning the mark point set by the movable camera. Based on the change of the position of the movable camera and the positioning information of the movable camera to the mark point set, the three-dimensional measurement information is spliced ​​to the initial position of the movable camera. While expanding the measurement range, only one coordinate conversion is added, ensuring high accuracy even in the case of large-scale measurement.

[0101] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be distributed to multiple network units. A person skilled in the art may select some or all of the modules according to actual needs to achieve the purpose of the solution of this embodiment without creative effort.

[0102] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A large-scale three-dimensional tracking method based on multi-camera fusion, characterized in that: The method comprises: A large-scale three-dimensional tracking system based on multi-camera fusion is constructed, the system comprising a servo motor, a movable camera, a search camera, a reference camera, a rotating stage, a reference calibration plate and a base; wherein the rotating stage is mounted on the base, the reference camera is fixedly connected to the base, the reference calibration plate is fixed under the movable camera, the servo motor and the search camera are fixed on the rotating stage, and the movable camera is fixedly connected to the rotating shaft of the servo motor; Calibrate the initial pose of the movable camera through the reference camera; The search camera tracks the landmark point set, calculates the rotation angle of the rotating stage and the servo motor, and the landmark point set is always placed within the field of view of the movable camera through the rotation of the rotating stage and the servo motor; The movable camera continuously measures the poses of the marker point sets at different measuring points. Meanwhile, the reference camera obtains the pose change information of the movable camera through the reference calibration plate under the movable camera. According to the pose information of the marker point set of different measuring points measured by the movable camera and the pose change information of the movable camera, the measurement results of the movable camera are stitched into the initial coordinate system of the movable camera; Before tracking the landmark point set by the search camera, the method further includes calibrating the relative position of the search camera and the movable camera, and calibrating the rotation axis of the rotating stage and the servo motor: The search camera and the movable camera are calibrated through an external calibration board to obtain the posture conversion relationship between the two: in, To search for the camera relative to the movable camera pose matrix, To search for the pose matrix of the camera relative to the reference calibration plate, is the pose matrix of the reference calibration plate relative to the movable camera; under this condition, the rotation angles of the rotating table and the servo motor are α0 and β0 respectively; When calibrating the rotating axis of the rotating stage and servo motor, the movable camera or search camera continuously takes pictures of an external calibration plate during the rotation of the rotating stage or servo motor, and obtains the l positions of the optical center of the movable camera or search camera in the coordinate system of the external calibration plate, and then calculates the rotating axis and a point on the rotating axis: n=(u,v,w) o=(a,b,c) Where n is the unit rotation axis vector, and o is a point on the rotation axis; According to the Rodriguez formula, the transformation formula of the rotation matrix R after the movable camera or the search camera rotates around the axis by an angle θ is obtained: R=cos θI+(1-cos θ)nn T +sin θn^ Where θ is the rotation angle, I is the identity matrix, n ∧ represents the antisymmetric matrix of n; according to the conversion formula, the rotation matrix R of the movable camera or the search camera rotating around the axis by an angle θ is obtained θ and the translation matrix t θ .

2. The large-scale three-dimensional tracking method based on multi-camera fusion according to claim 1 is characterized in that: The initial position and posture of the movable camera is calibrated by the reference camera and specifically includes: The reference camera calibrates the initial position and posture of the movable camera by taking pictures of a reference calibration plate fixedly connected to the movable camera.

3. The large-scale three-dimensional tracking method based on multi-camera fusion according to claim 1 is characterized in that: The method of tracking the marker point set by searching the camera, calculating the rotation angle of the rotating table and the servo motor, and placing the marker point set always within the field of view of the movable camera by rotating the rotating table and the servo motor specifically includes: The rotating stage drives the search camera to rotate until the set of landmark points appears in the field of view of the search camera; Calculate the centroid of the marker point set and obtain the coordinates of the centroid of the marker point set in the search camera coordinate system, recorded as (x, y, z); According to the calibrated rotation axis of the rotating stage and servo motor and the position conversion relationship between the search camera and the movable camera, the coordinates (x′, y′, z′) of the center of gravity of the marker point in the initial position coordinate system of the movable camera are obtained as follows: Where α is the rotation angle of the rotating table relative to α0, is the transformation matrix of the movable camera relative to the initial position after the rotating table rotates around the axis α based on the angles of the rotating table and the servo motor being α0 and β0, R α is the corresponding rotation matrix, t α is the corresponding translation matrix; To search for the pose matrix of the camera relative to the reference calibration plate, is the pose matrix of the reference calibration plate relative to the movable camera; The rotation angles of the rotating table and the servo motor are calculated respectively, and the rotation of the rotating table and the servo motor is controlled according to the rotation angles so that the center of gravity of the marker point set is directly opposite to the optical axis of the movable camera.

4. The large-scale three-dimensional tracking method based on multi-camera fusion according to claim 3 is characterized in that: The steps of respectively calculating the rotation angles of the rotating table and the servo motor, and controlling the rotating table and the servo motor to rotate according to the rotation angles so that the center of gravity of the marker point set faces the optical axis of the movable camera specifically include: Assuming that the rotation angles of the rotating table and servo motor are α′ and β′ respectively, we can get: in, R α′ ,t α′ They are the rotation matrix and translation matrix corresponding to the rotation angle of the rotating table of α′, R β′ ,t β′ They are the rotation matrix and translation matrix corresponding to the rotation angle of the servo motor of β′. The above formula is simplified into a set of two-variable linear equations about α′ and β′: Among them, S1-S8 are known constants. Solve this set of two-variable linear equations to obtain the values ​​of α′ and β′; The rotating table and the servo motor are controlled to rotate according to the values ​​of the rotation angles α′ and β′ so that the center of gravity of the marker point set is directly opposite to the optical axis of the movable camera.

5. The large-scale three-dimensional tracking method based on multi-camera fusion according to claim 1, characterized in that: According to the pose information of the marker point set of different measuring points measured by the movable camera and the pose change information of the movable camera, the measurement results of the movable camera are stitched into the initial pose of the movable camera, which specifically includes: Suppose that the m poses of the landmark point set of different measuring points are obtained by continuous measurement by a movable camera, i=1,2,…,m, m is the total number of measurement points in the process of moving the marker point set. The reference camera measures the reference calibration plate fixed to the movable camera and obtains the pose change of the movable camera relative to the initial pose as ΔT i , then the pose information is stitched to the initial pose: The three-dimensional measurement results of the movable camera in the initial coordinate system are obtained.

6. The large-scale three-dimensional tracking method based on multi-camera fusion according to claim 1, characterized in that: The movable camera, search camera and reference camera are all monocular cameras.

7. The large-scale three-dimensional tracking method based on multi-camera fusion according to claim 6 is characterized in that the set of marker points is a light pen or other measuring equipment carrying three-dimensional information.