A mechanical arm and a three-dimensional point cloud splicing method based on the mechanical arm

Abstract

The application discloses a kind of mechanical arm and three-dimensional point cloud splicing method based on mechanical arm, it is related to three-dimensional scanning field, it includes, S1: setting calibration scanning path;S2: calibration picture acquisition;S3: calibration three-dimensional scanner parameter;S4: setting model reconstruction scanning path;S5: acquisition point cloud data;S6: two side point cloud splicing fusion;S7: multi-view point cloud splicing fusion.The three-dimensional point cloud splicing method based on mechanical arm provided in the application is simple to operate, just needs to calibrate the relationship between each scanning position in advance to complete the splicing of point cloud data, without human intervention, also does not need to place mark point on scanning model, and does not need to calibrate between mechanical arm and three-dimensional scanner, so that the splicing precision of point cloud model will not be influenced by mechanical arm DH parameter, which improves the stability and accuracy of point cloud splicing.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of three-dimensional scanning, in particular to a mechanical arm and a three-dimensional point cloud splicing method based on the mechanical arm. BACKGROUND

[0002] It is known that three-dimensional cloud splicing refers to splicing and fusing multiple three-dimensional data using cloud computing technology to generate a more complete and accurate three-dimensional model. Through three-dimensional cloud splicing, scattered three-dimensional data can be combined into a whole, thereby better understanding and analyzing information in three-dimensional space.

[0003] For example, a patent document with the name "Three-dimensional point cloud splicing method based on standard ball stand" and the authorization date of April 7, 2023 (CN113269673B) uses a standard ball stand as an intermediate medium to unify each local coordinate system in the measurement field of view area, eliminating the cumulative error caused by multi-field-of-view splicing. Meanwhile, in the process of point cloud accurate registration, a virtual overlap area is constructed using a spherical point cloud with feature constraints, providing more accurate corresponding point pairs. In combination with an improved iterative closest point algorithm, a weight function is used to optimize the solution of the transformation relationship between spliced point clouds, reducing the influence of large noise in the point cloud, and achieving high-precision three-dimensional point cloud splicing.

[0004] The deficiencies of the prior art are that there are mainly two methods for three-dimensional point cloud data splicing based on a mechanical arm, namely a splicing method based on a marker point and a splicing method based on robot kinematics. The splicing method based on a marker point needs to place a marker point on a model, so the splicing accuracy depends on the extraction accuracy of the marker point. If the marker point is not placed in place, it will seriously affect the accuracy of point cloud splicing. In addition, the marker point needs to be placed again every time a model is scanned in the process of this method, which is very tedious. In comparison, the splicing method based on robot kinematics does not need to place a marker point. It converts the point clouds of multiple fields of view from the coordinate system of the scanner to a common robot coordinate system, thereby realizing point cloud data splicing. The splicing accuracy depends on the accuracy of the calibration parameters of the scanner, the hand-eye calibration parameters, and the DH parameters of the robot. Therefore, this method will inevitably introduce mechanical arm model errors. When the positioning accuracy of the mechanical arm is very low, the splicing accuracy of this method is very unsatisfactory. SUMMARY

[0005] The purpose of the present application is to provide a mechanical arm and a three-dimensional point cloud splicing method based on the mechanical arm to solve the above-mentioned deficiencies in the prior art.

[0006] In order to achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0007] A three-dimensional point cloud splicing method based on a mechanical arm,

[0008] S1: setting a calibration scanning path;

[0009] S2: acquiring calibration pictures;

[0010] S3: calibrating three-dimensional scanner parameters;

[0011] S4: setting a scanning path for model reconstruction;

[0012] S5: acquiring point cloud data;

[0013] S6: splicing and fusing point clouds on two sides;

[0014] S7: splicing and fusing point clouds in multiple perspectives.

[0015] The three-dimensional point cloud splicing method based on a mechanical arm, in step S1, is based on a structure of two cameras and a projector. A model is placed in a specified placement area, and the projector projects a beam of light toward the model. The positions of the mechanical arm are adjusted based on the calibration board pictures taken by the two cameras, so as to determine multiple positions of the three-dimensional scanner calibration, and record the mechanical arm positions of each calibration position for subsequent camera and projector calibration.

[0016] The three-dimensional point cloud splicing method based on a mechanical arm, in step S2, first communicates the mechanical arm with the three-dimensional scanner. After successful communication, the three-dimensional scanner is calibrated, and then the mechanical arm is controlled to move according to the set calibration positions. After moving to the specified position, the mechanical arm stops, and waits for the three-dimensional scanner to take the calibration pictures of the current position before moving to the next calibration position. After all positions are completed, the mechanical arm returns to the zero position.

[0017] The three-dimensional point cloud splicing method based on a mechanical arm, in step S3, calibrates the intrinsic matrix, distortion coefficient matrix and translation matrix of the camera and projector based on the calibration pictures taken by the three-dimensional scanner, and also calibrates the rotation and translation relationship between the two cameras and between the two cameras and the projector, which is used for subsequent point cloud splicing on the left and right sides.

[0018] The three-dimensional point cloud splicing method based on a mechanical arm, in step S4, controls the movement of the mechanical arm so that the three-dimensional scanner can realize omnidirectional scanning of the model, and finally calibrates the relative position relationship between each scanning position.

[0019] The three-dimensional point cloud splicing method based on a mechanical arm, in step S5, when the mechanical arm communicates successfully with the three-dimensional scanner, the mechanical arm moves along the set scanning path. When it reaches a position, the three-dimensional scanner scans the model. After scanning is completed, the mechanical arm moves to the next position. After all positions are completed, the three-dimensional point cloud data of the model at each position is obtained.

[0020] The three-dimensional point cloud splicing method based on a mechanical arm, in step S6, two cameras and a projector form two monocular structured light systems respectively, two point cloud data can be constructed at each point, and the two point cloud data are obtained from two perspectives of the model, and before splicing the point cloud from multiple perspectives, the two point cloud data corresponding to each point need to be fused to obtain a complete point cloud data.

[0021] The three-dimensional point cloud splicing method based on a mechanical arm, in step S7, after obtaining the complete point cloud data under each perspective, the three-dimensional point cloud data under each perspective can be unified to the same perspective according to the relative position relationship between the calibrated scanning positions, so as to obtain a complete three-dimensional point cloud model.

[0022] A mechanical arm for driving a three-dimensional scanner to scan a model with an internal hollow structure, characterized in that it comprises a mechanical arm body, the three-dimensional scanner is arranged at the execution end of the mechanical arm body, and further comprises a fine adjustment mechanism for adjusting the rotation angle and axial movement between the three-dimensional scanner and the execution end of the mechanical arm body.

[0023] The mechanical arm, the fine adjustment mechanism comprises a gas cylinder, an installation groove is formed in the mechanical arm body, the gas cylinder is arranged in the installation groove, a rotating seat is slidably arranged on the mechanical arm body, the rotating seat is connected with the output end of the gas cylinder, a rotating shaft is rotatably arranged on the rotating seat, and the three-dimensional scanner is fixedly connected to the rotating shaft.

[0024] In the above technical solution, the three-dimensional point cloud splicing method based on a mechanical arm provided by the application is simple to operate, and the splicing of point cloud data can be completed only by calibrating the relationship between the scanning positions in advance, without human intervention, without the need to place a marker on the scanned model, and without the need to calibrate the mechanical arm and the three-dimensional scanner, so that the splicing accuracy of the point cloud model is not affected by the DH parameters of the mechanical arm, thereby improving the stability and accuracy of point cloud splicing. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0026] Figure 1 The flowchart of the splicing method provided by the embodiments of the present application;

[0027] Figure 2This is a schematic diagram of the external structure provided for another embodiment of the present invention;

[0028] Figure 3 This is a side view structural schematic diagram provided for another embodiment of the present invention;

[0029] Figure 4 This is a partial cross-sectional structural schematic diagram provided in another embodiment of the present invention;

[0030] Figure 5 This is a schematic diagram of the overall structure of the cylindrical segment provided in another embodiment of the present invention;

[0031] Figure 6 for Figure 2 Enlarged schematic diagram of a local structure at point A;

[0032] Figure 7 for Figure 4 Enlarged schematic diagram of the local structure at point B;

[0033] Figure 8 for Figure 4 A magnified schematic diagram of the local structure at point C.

[0034] Explanation of reference numerals in the attached figures:

[0035] 1. Robotic arm body; 2. 3D scanner; 3. Cylinder; 4. Mounting slot; 5. Rotating seat; 6. Rotating shaft; 7. Guide rod; 8. Drive motor; 9. Sliding seat; 10. Transmission frame; 11. First wedge block; 12. Second wedge block; 13. First spring; 14. Locking block; 1401. Locking section; 1402. Cylindrical section; 15. Locking groove; 16. Rotating hole; 17. Limiting cylinder; 18. Limiting groove; 19. Second spring; 20. First abutting block; 2001. Abutting section; 2002. Sliding section; 21. Second abutting block; 22. Third spring; 23. Locking rod; 24. Locking hole; 25. Fourth spring; 26. Conducting channel; 27. Push plate; 28. Drive block; 29. ​​Collection groove; 30. Fifth spring; 31. Rotating plate; 32. Output shaft; 33. Drive groove. Detailed Implementation

[0036] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0037] In the description of this invention, it should be understood that... Figure 3 The axis of the middle cylinder 3 is vertical. Figure 3The width direction of the three-dimensional scanner 2 is the horizontal direction, and the terms "center", "longitudinal", "width", "degree", "upper", "lower", "vertical", "horizontal", "top", "bottom", "inner", "outer", "axial", "radial", "circumferential", and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0038] Referring to Figures 1-8 The three-dimensional point cloud splicing method based on a mechanical arm provided by the embodiments of the present application,

[0039] S1: setting a calibration scanning path;

[0040] S2: acquiring calibration pictures;

[0041] S3: calibrating the three-dimensional scanner 2 parameters;

[0042] S4: setting a model reconstruction scanning path;

[0043] S5: acquiring point cloud data;

[0044] S6: two-side point cloud splicing and fusion;

[0045] S7: multi-view point cloud splicing and fusion.

[0046] Specifically, three-dimensional point cloud stitching is a process of integrating multiple three-dimensional point cloud data at different positions or angles to form a complete three-dimensional model. Common three-dimensional point cloud stitching methods include: first, feature matching and registration. Feature extraction and matching are first performed on different three-dimensional point cloud data to find the corresponding relationship between them, and then registration is performed, that is, they are aligned to the same coordinate system. Second, stitching based on geometric information. Different point cloud data are stitched and fused using geometric information of the point cloud, such as normal vector and curvature. Common methods include ICP (Iterative Closest Point) algorithm and stitching algorithm based on geometric features. Third, stitching based on deep learning. Deep learning methods are used to learn and match features of different point cloud data to achieve automatic stitching of point clouds. This is prior art and will not be described in detail. One of the core innovations of the embodiments of the present application is that when stitching point clouds, the following steps are performed: S1: based on a structure of two cameras and a projector, the model is placed in a designated placement area, the projector projects a beam of light onto the model, and the positions of the three-dimensional scanner 2 are determined based on the calibration board pictures taken by the two cameras, and the point positions of the mechanical arm at each calibration position are recorded for subsequent camera and projector calibration; S2: calibration picture acquisition. First, the mechanical arm communicates with the three-dimensional scanner 2. After successful communication, the three-dimensional scanner 2 is calibrated, and then the mechanical arm moves according to the set calibration positions. After moving to the designated position, the mechanical arm stops, and after the three-dimensional scanner 2 finishes shooting the calibration picture at the current position, it moves to the next calibration position. After all positions are completed, the mechanical arm returns to the zero position; S3: calibrate the parameters of the three-dimensional scanner 2. Based on the calibration pictures taken by the three-dimensional scanner 2, the intrinsic matrix, distortion coefficient matrix and translation matrix of the camera and projector are calibrated, and the rotation and translation relationship between the two cameras and between the two cameras and the projector is also calibrated for subsequent left and right point cloud stitching; S4: set the scanning path for model reconstruction, control the mechanical arm to move, so that the three-dimensional scanner 2 can realize full-range scanning of the model, and finally calibrate the relative position relationship between each scanning position; S5: acquire point cloud data. After successful communication between the mechanical arm and the three-dimensional scanner 2, the mechanical arm moves along the set scanning path. When it reaches a position, the three-dimensional scanner 2 scans the model. After scanning is completed, the mechanical arm moves to the next position. After all positions are completed, the three-dimensional point cloud data of the model at each position is obtained; S6: left and right point cloud stitching and fusion. Two cameras and a projector form two monocular structured light systems respectively. Two point cloud data can be constructed at each point. Before stitching the multi-view point cloud, the two point cloud data corresponding to each point need to be fused to obtain a complete point cloud data.S7: Multi-view point cloud splicing and fusion, after obtaining complete point cloud data under each view, according to the relative position relationship between each scanning position calibrated, the three-dimensional point cloud data under each view can be unified to the same view, so as to obtain a complete three-dimensional point cloud model, the advantages of this method are: first, only the relationship between each scanning position needs to be calibrated in advance to complete the splicing of point cloud data, without human intervention, second, without placing a marker on the scanning model, and without calibrating between the mechanical arm and the three-dimensional scanner 2, third, the splicing accuracy of the point cloud model will not be affected by the DH parameters of the mechanical arm, which improves the stability and accuracy of the point cloud splicing.

[0047] The specific process is as follows: first, since the three-dimensional scanner 2 used in the present application is a dual-camera, single-projector structure, it is necessary to combine the point clouds reconstructed by the left and right monocular structured light systems in the same camera coordinate system to obtain complete point cloud data of the model under a single view. Taking the left monocular structured light system as an example, the equation of the coordinates (X w ,Y w ,Z w ) of a point in the world coordinate system is as follows:

[0048]

[0049] In the above formula, K c and K p represent the intrinsic matrices of the camera and the projector respectively, R c and T c , R p and T p represent the extrinsic matrices of the camera and the projector respectively, these parameters are obtained during the calibration of the camera and the projector. By solving the above equation, the coordinates of a point on the model in the world coordinate system can be obtained, and by making the world coordinate system coincide with the camera coordinate system, the above equation will become:

[0050]

[0051] In the above formula, since the world coordinate system coincides with the camera coordinate system, there is no rotation and translation relationship between the two coordinate systems, i.e.:

[0052]

[0053] And the rotation and translation relationship R p and T p between the projector coordinate system and the world coordinate system becomes the rotation and translation relationship R cp and T cp between the camera coordinate system, i.e.:

[0054]

[0055] Secondly, after the point clouds reconstructed by the left and right monocular structured light systems are unified to the camera coordinate system, the next step is to fuse the point clouds of the left and right sides. By stereo calibration of the two cameras, the relative position parameters R and T between the two cameras can be obtained, and through R and T, the right point cloud P r can be converted to the camera coordinate system of the left point cloud P l . The conversion formula is as follows:

[0056] P l = RP r +T

[0057] After the right point cloud data is converted to the left camera coordinate system, in order to avoid data redundancy, the repeated parts in the point cloud data of the left and right sides need to be deleted, and only the parts different from the left point cloud in the right point cloud are superimposed on the left point cloud.

[0058] Then, after fusing the left and right point clouds, the next step is to obtain the point cloud data of the model at each scanning position according to the set scanning path, and then splice and fuse the point cloud data at each viewing angle. The scanning positions set by the application are preferably the upper, front, left front, left, left rear, rear, right rear, right, and right front positions of the model. Taking the nine positions as an example, by keeping the model stationary and sequentially obtaining the point cloud data of the model at each scanning position in the above order, the three-dimensional point cloud data at each position can be unified to the initial position by solving the rotation and translation relationship between each position and the initial position. The equation is as follows:

[0059]

[0060] In the above formula, P c0 represents the coordinates of point P at the initial scanning position in the camera coordinate system, P ci represents the coordinates of point P at the i-th scanning position in the camera coordinate system, where i=0, 1, 2, …, n, P w represents the coordinates of point P in the world coordinate system, R c0 and T c0 , and R ci and T ci can be solved by PNP algorithm. Only by pre-calibrating the external parameter matrix between the camera coordinate system and the world coordinate system at each position, the rotation and translation relationship R i and T i between each position and the initial position can be obtained, that is:

[0061]

[0062] M i =[Ri T i ]

[0063] Assuming that the model point cloud data obtained at each scanning position is P i , where i represents 9 viewing angles, the 0th position, i.e. the position above the model, is taken as the initial position, and the point clouds of the remaining 8 viewing angles are coarsely spliced by the rotation and translation matrix obtained in step 3, i.e.

[0064] P0' = M0P0

[0065] P1' = M1P1 ...

[0067] P8' = M8P8

[0068] At this time, all the model point cloud data is located under the viewing angle of the initial position, and the point cloud coarse splicing process is completed.

[0069] Finally, after the point cloud coarse splicing is completed, the point cloud models at each position are basically spliced together, but the splicing accuracy is not high. In order to further improve the accuracy of the point cloud splicing, ICP fine splicing needs to be performed on the coarsely spliced point cloud to improve the overall splicing accuracy of the three-dimensional point cloud model.

[0070] In the above embodiments, the mechanical arm itself can theoretically realize a certain range of the present application. Another embodiment of the present application provides a mechanical arm for driving a three-dimensional scanner 2 to realize scanning of a model with an internal hollow structure, characterized in that it comprises a mechanical arm body 1, the three-dimensional scanner 2 is arranged at the execution end of the mechanical arm body 1, and further comprises a fine adjustment mechanism for adjusting the rotation angle and axial movement between the three-dimensional scanner 2 and the execution end of the mechanical arm body.

[0071] Specifically, the plurality of joints of the mechanical arm body 1 can rotate relative to each other, thereby achieving positioning of the execution end at any position and any angle within the movement range. The three-dimensional scanner 2 is installed at the execution end of the mechanical arm body 1, and the position and angle of the three-dimensional scanner 2 can be adjusted by rotating the plurality of joints of the mechanical arm body 1. After the three-dimensional scanner 2 has passed all the calibration positions, the external three-dimensional graphics of the model can be obtained by splicing the point cloud model captured by the three-dimensional scanner 2. This is prior art and will not be described in detail. However, in the prior art, only the point cloud data of the outer surface of the model can be collected. When it is necessary to obtain the point cloud data inside a model (such as a box or a barrel, hereinafter referred to as a model) having a large recess or even a hollow interior, the three-dimensional scanner 2 cannot be extended into the interior of the model or observe a large recess due to the length limitation of each joint of the mechanical arm body 1 and the inability of the three-dimensional scanner 2 to move relative to the execution end, and the rotary connection between each joint. Even if a small part can be extended into the interior of the model, the rotation of the joint when adjusting the angle of the three-dimensional scanner 2 is likely to interfere with the entry side wall of the model, so that complete point cloud data inside the model cannot be obtained. One of the core innovations of the embodiment of the present application is that a fine adjustment mechanism is arranged between the mechanical arm body 1 and the three-dimensional scanner 2. The fine adjustment mechanism is a reciprocating drive assembly such as an electric push rod and a drive assembly that provides rotational force for the motor. It is used to adjust the angle of the three-dimensional scanner 2 and the distance between the three-dimensional scanner 2 and the mechanical arm body 1. The purpose of such arrangement is that when it is necessary to obtain point cloud data inside the model, power is provided by the fine adjustment mechanism, so that the three-dimensional scanner 2 itself can be extended to the inner bottom wall of the model, and then the shooting angle of the three-dimensional scanner 2 is adjusted by the fine adjustment mechanism, so that the point cloud data of the interior space of the model can be obtained in all directions. Under normal circumstances, the fine adjustment mechanism does not intervene, and only when the model has a large recess or hollow structure does it intervene.

[0072] Preferably, the fine adjustment mechanism comprises a cylinder 3, a mounting groove 4 is formed on the execution end of the mechanical arm body 1, the cylinder 3 is arranged in the mounting groove 4, a rotating seat 5 is also slidingly arranged on the execution end of the mechanical arm body 1, the rotating seat 5 is connected with the output end of the cylinder 3, a rotating shaft 6 is rotatably arranged on the rotating seat 5, the three-dimensional scanner 2 is fixedly connected with the rotating shaft 6, specifically, the mounting groove 4 is formed on the execution end of the mechanical arm body 1, the cylinder 3 is arranged in the mounting groove 4 and coaxially arranged with the mounting groove 4, the rotating seat 5 is in the shape of a rectangle, a guide rod 7 is arranged on the side wall of the rotating seat 5, a guide groove is formed on the execution end of the mechanical arm body 1 and matched with the guide rod 7, the guide rod 7 is slidingly connected with the guide groove, the output end of the cylinder 3 is connected with the upper surface of the rotating seat 5, a through hole is formed on the rotating seat 5 and matched with the rotating shaft 6, the rotating shaft 6 is rotatably connected with the through hole, and a limiting structure (not shown in the figure) for preventing axial movement of the two is arranged between the rotating shaft 6 and the through hole, a connecting part is arranged on the upper surface of the three-dimensional scanner 2 and fixedly connected with the rotating shaft 6, the arrangement has the following effects: when the point cloud data of the inside or recessed part of the model is needed, the three-dimensional scanner 2 is moved to the hollow entrance of the model by the mechanical arm body 1, the cylinder 3 is coaxially arranged with the model entrance, then the driving rod of the cylinder 3 is controlled to be elongated, so as to drive the guide rod 7 to slide with the guide groove, so that the rotating seat 5 and the three-dimensional scanner 2 can be transported to the inside of the model, then the rotation of the mechanical arm body 1 and the rotating shaft 6 can control the rotation of the three-dimensional scanner 2, so that the complete point cloud data of the inside of the model can be obtained.

[0073] Preferably, the fine adjustment mechanism further comprises a driving motor 8, the driving motor 8 is slidingly arranged on the rotating seat 5, the driving motor 8 has a first position connected with the rotating shaft 6 and a second position disconnected with the rotating shaft 6; further comprising a power assembly for driving the driving motor 8 to switch between the first position and the second position; further comprising a locking assembly for fixing the rotating shaft 6 when the driving motor 8 is in the second position, specifically, the rotating seat 5 is provided with a sliding seat 9, the driving motor 8 is slidingly connected with the sliding seat 9 through the base, the rotating shaft 6 is provided with a driving groove 33 matched with an output shaft 32 of the driving motor 8, and a protruding structure for preventing relative rotation between the two is arranged between the two, the power assembly is a reciprocating moving assembly such as an electric push rod, and the locking assembly is a matching structure of a bolt and a slot. The setting has the following effects: when the output end of the driving motor 8 is inserted into the driving groove 33, the driving motor 8 can be connected with the rotating shaft 6, so that the angle of the rotating shaft 6 and the three-dimensional scanner 2 can be adjusted. When the point cloud data outside the model is normally acquired, the mechanical arm body 1 and the cylinder 3 are sufficient to acquire the point cloud data outside the model, at this time, the driving motor 8 does not need to work, and the output end of the cylinder 3 does not need to be elongated. In order to make the output shaft 32 of the driving motor 8 in a stress-free state when the driving motor 8 is not working, when the driving motor 8 is not used, the driving motor 8 is switched from the first position to the second position by the power assembly, so that the output shaft 32 of the driving motor 8 is away from the driving groove 33, so as to avoid the driving motor 8 in a stressed state, thereby improving the stability of the three-dimensional scanner 2 and prolonging the service life of the driving motor 8. At the same time, the locking assembly is started, so that the rotating shaft 6 can be locked synchronously when the driving motor 8 is switched in position, so as to avoid the rotating shaft 6 from rotating by itself after losing the driving force of the driving motor 8.

[0074] As a further embodiment of the present application, the power assembly comprises a transmission frame 10 fixedly connected to the base of the driving motor 8, a first wedge-shaped block 11 fixedly connected to the transmission frame 10, and a second wedge-shaped block 12 fixedly connected to the mechanical arm body 1, wherein the first wedge-shaped block 11 is located in the movement stroke of the second wedge-shaped block 12, and the transmission frame 10 is provided with a first spring 13 between the transmission frame 10 and the rotating seat 5. Specifically, the transmission frame 10 is approximately U-shaped, one end of which is fixedly connected to the housing of the driving motor 8, and the other end is located on one side of the rotating shaft 6. The first wedge-shaped block 11 is fixedly connected above the horizontal section of the transmission frame 10, and the second wedge-shaped block 12 is fixedly connected to the lower surface of the execution end of the mechanical arm body 1. The arrangement is to enable the point cloud data outside the model to be obtained without the need for the cylinder 3 and the driving motor 8 to work, that is, the output end of the cylinder 3 is controlled to be retracted. During the retraction of the output end of the cylinder 3, the rotating seat 5 and the transmission frame 10 are synchronously moved. When the wedge-shaped surfaces of the first wedge-shaped block 11 and the second wedge-shaped block 12 abut, the wedge-shaped cooperation drives the transmission frame 10 to move horizontally, that is, to move in the axial direction of the rotating shaft 6, so that the output shaft 32 of the driving motor 8 and the driving groove 33 are passively separated, and the first spring 13 is stretched. The output end of the cylinder 3 is manually or electrically locked by existing structures such as bolts or pins to avoid the output end of the cylinder 3 being always in a stressed state. When the point cloud data inside the model needs to be obtained, that is, the output end of the cylinder 3 needs to be controlled to be elongated, and the output shaft 32 of the driving motor 8 needs to be inserted into the driving groove 33 again. At this time, the locking function of the cylinder 3 is released, and then the output end of the cylinder 3 is controlled to be elongated, which causes the first wedge-shaped block 11 and the second wedge-shaped block 12 to move away from each other. The driving motor 8 is automatically reset under the elastic force of the first spring 13, so that the output shaft 32 of the driving motor 8 is passively inserted into the driving groove 33 to realize the transmission of kinetic energy.

[0075] Preferably, the locking assembly comprises a locking block 14 arranged on the transmission frame 10, and the rotating shaft 6 is provided with a locking groove 15 matched with the locking block 14. When the driving motor 8 is in the first position, the locking block 14 is away from the locking groove 15. When the driving motor 8 is in the second position, the locking block 14 is inserted into the locking groove 15. Specifically, the cross section of the locking block 14 is preferably a regular polygon such as an octagon, which is arranged at the end of the transmission frame 10 away from the driving motor 8. The locking groove 15 is coaxially arranged with the rotating shaft 6 and is also a regular polygon such as an octagon. When the driving motor 8 is in the first position, the distance between the locking block 14 and the locking groove 15 is less than the length of the output end of the driving motor 8 inserted into the driving groove 33. The arrangement is to ensure that, when the transmission frame 10 moves with the driving motor 8 during the switching of the driving motor 8 from the first position to the second position, the locking block 14 moves synchronously, so that the locking block 14 is inserted into the locking groove 15. Since the distance between the locking block 14 and the locking groove 15 is less than the length of the output end of the driving motor 8 inserted into the driving groove 33, the output shaft 32 of the driving motor 8 is extracted from the driving groove 33 after the locking block 14 is inserted into the locking groove 15, so that the locking function of the locking block 14 and the fixing function of the driving motor 8 are seamlessly connected. Thus, after the driving motor 8 is switched from the first position to the second position, the position of the rotating shaft 6 and the three-dimensional scanner 2 can still be fixed under the insertion of the locking block 14 and the locking groove 15.

[0076] When the drive motor 8 stops, if the corner of the locking block 14 is not aligned with the corner of the locking groove 15, the locking block 14 may not be able to be inserted into the locking groove 15. In another embodiment of the invention, the transmission frame 10 has a rotating hole 16 inside, and a limiting cylinder 17 is fixedly connected to the inner wall of the rotating hole 16. The outer circumferential surface of the locking block 14 has a limiting groove 18 that matches the limiting cylinder 17. The limiting cylinder 17 and the limiting groove 18 are slidably connected, and a second spring 19 is provided between the locking block 14 and the side wall of the rotating hole 16. Specifically… The locking block 14 includes an integral locking section 1401 and a cylindrical section 1402. The locking section 1401 has a regular octagonal cross-section. A rotating plate 31 is provided on the side wall of the rotating hole 16. The two ends of the second spring 19 are fixedly connected to the rotating plate 31 and the cylindrical section 1402, respectively. The limiting cylinder 17 is cylindrical, and the limiting groove 18 is arranged along the circumference of the cylindrical section 1402 and is arc-shaped. This arc extends in both the circumference and axial direction of the cylindrical section 1402, or more precisely, it is part of a spiral. The purpose of this arrangement is to allow the drive motor 8 to switch from the first position to the second position. When positioned, the transmission frame 10 will cause the locking block 14 to move towards the locking groove 15. If the corner of the locking section 1401 is not aligned with the corner of the locking groove 15, the locking block 14 will not be able to be inserted into the locking groove 15. At this time, the locking section 1401 will abut against the side wall of the rotating shaft 6, causing the cylindrical section 1402 to contract towards the inside of the rotating hole 16 and compress the second spring 19. During the contraction process, the limiting cylinder 17 will slide along the limiting groove 18. Since the limiting groove 18 is arc-shaped, the cylindrical section 1402 and the locking section 1401 will slide and rotate simultaneously (rotation). The angle is between 0 and 45 degrees. Under the action of the second spring 19, the rotating plate 31 is driven to rotate synchronously. When the locking section 1401 rotates to the point where its edge is aligned with the edge of the locking groove 15, the locking block 14 will be inserted into the locking groove 15 to achieve the passive correction function. When the drive motor 8 switches from the second position to the first position, it will drive the transmission frame 10 to move in the opposite direction. When the transmission frame 10 moves, it will drive the locking block 14 to be pulled out from the locking groove 15. At this time, the elastic force of the second spring 19 is released, and the locking block 14 automatically resets under the action of the second spring 19.

[0077] After the locking block 14 is inserted into the locking groove 15, the rotating shaft 6 is not completely locked because the positions of the limiting cylinder 17 and the limiting groove 18 are not locked. When the rotating shaft 6 swings slightly, it will cause the cylindrical section 1402 to slide relative to the rotating hole 16. Furthermore, a first abutting block 20 is slidably arranged on the transmission frame 10, and a third spring 22 is arranged between the first abutting block 20 and the transmission frame 10. A locking rod 23 is elastically arranged on the first abutting block 20. A locking hole 24 adapted to the locking rod 23 is opened on the inner wall of the limiting groove 18. A second abutting block 21 is fixedly connected to the rotating seat 5. The first abutting block 20 is located on the movement stroke of the second abutting block 21. Specifically, the first abutting block 20 is located on the movement stroke of the second abutting block 21. Both the first abutment block 20 and the second abutment block 21 are provided with wedge-shaped surfaces. The first abutment block 20 includes an abutment section 2001 and a sliding section 2002. The cross-section of the sliding section 2002 is square (to prevent the first abutment block 20 from rotating itself when sliding). The limiting cylinder 17 is a hollow cylindrical structure. The sliding section 2002 slides with both the rotating frame and the space inside the limiting cylinder 17. The wedge-shaped surface is provided at the top of the abutment section 2001. The third spring 22 is provided between the abutment section 2001 and the transmission frame 10. The elastic setting, i.e., the locking rod 23, is provided at the bottom of the sliding section 2002, and a fourth spring 25 is provided between it and the sliding section 2002. Multiple locking holes 24 are provided and are evenly arranged along the direction of the limiting groove 18. The function is that when the transmission frame 10 drives the locking block 14 to move towards the locking groove 15, the wedge-shaped surface on the abutting section 2001 abuts against the wedge-shaped surface on the second abutting block 21, thereby driving the abutting section 2001 and the sliding section 2002 to move vertically downward and compress the third spring 22, thereby driving the locking rod 23 to move downward synchronously. During the downward movement of the locking rod 23, there are two situations. First, when the locking rod 23 is aligned with one of the locking holes 24, the locking rod 23 will be inserted into the locking hole 24, making it impossible for the locking rod 23 and the locking hole 24 to move relative to each other in the horizontal direction. Since the locking rod 23 is located inside the limiting cylinder 17, the limiting cylinder 17 cannot continue to move with the limiting groove 18. The cylindrical segment 1402 and the rotating hole 16 cannot slide relative to each other, thus passively locking the position of the locking block 14, thereby fixing the positions of the rotating shaft 6 and the 3D scanner 2. Secondly, when the locking rod 23 is not aligned with the locking hole 24, the bottom end of the locking rod 23 will abut against the bottom wall of the limiting groove 18 and compress the fourth spring 25. At this time, the cylindrical segment 1402 is not completely fixed and will slide slightly inside the rotating hole 16. When the cylindrical segment 1402 slides, it will drive the locking holes 24 to slide synchronously. When one of the locking holes 24 is aligned with the locking rod 23, the elastic force of the fourth spring 25 is released, thereby allowing the locking rod 23 to be inserted into the locking groove 15.In this way, locking block 14 and locking groove 15 can be completely locked. When the transmission frame 10 moves locking block 14 away from locking groove 15, locking block 14 will be pulled out of locking groove 15, and the first abutting block 20 and the second abutting block 21 will move away from each other. At this time, the abutting effect of the second abutting block 21 on the first abutting block 20 disappears, and the first abutting block 20 will automatically reset under the elastic force of the third spring 22.

[0078] It should be noted that when switching the position of drive motor 8, the three-dimensional level needs to be adjusted to a horizontal state, and at this time, cylinder 3 should be in a vertical state (e.g., Figure 3 That is, the 3D scanner 2 and the cylinder 3 are perpendicular to each other.

[0079] During prolonged use of this robotic arm, especially in dusty production environments, dust will accumulate inside the locking groove 15. A collection groove 29 is provided inside the locking groove 15. A transmission channel 26 is provided inside the rotating shaft 6. A push plate 27 and a drive block 28 are dynamically sealed at both ends of the transmission channel 26. The collection groove 29 is located along the travel stroke of the push plate 27. A fifth spring 30 is provided between the drive block 28 and the side wall of the transmission channel 26. The stiffness coefficient of the fifth spring 30 is less than that of the second spring 19. Specifically, the collection groove 29 is located at the inner bottom of the locking groove 15. The wall is not interfered with by the locking block 14. The transmission channel 26 is U-shaped. The drive block 28 is coaxially arranged with the rotating shaft 6 and is located on the movement stroke of the locking block 14. The cross-sectional area of ​​the drive block 28 is larger than that of the push plate 27. The transmission channel 26 is filled with a transmission medium such as air. Protrusions or other structures are provided between the drive block 28 and the transmission channel 26, and between the push plate 27 and the transmission channel 26 to prevent them from separating. This is prior art and will not be described in detail. The purpose of this arrangement is that when the position of the drive motor 8 is switched, the 3D scanner 2 is in a horizontal state, and at this time the collection slot 29 is located in the locking slot. The inner bottom wall of the locking groove 15 allows dust and other impurities inside the locking groove 15 to slide down through the side wall of the locking groove 15 into the collection groove 29. When the locking block 14 slides towards the locking groove 15, the locking section 1401 abuts against the end of the driving block 28, thereby pushing the driving block 28 to slide into the conduction channel 26, simultaneously squeezing the fifth spring 30 and the conduction medium inside the conduction channel 26. Under the action of the conduction medium, the push plate 27 is squeezed, causing the push plate 27 to slide inside the collection groove 29. Since the cross-sectional area of ​​the driving block 28 is larger than the cross-sectional area of ​​the push plate 27, the push plate 27 moves... The distance the pusher 27 moves is greater than the distance the drive block 28 moves, so that the pusher 27 can completely push out the dust inside the collection groove 29. This passively pushes out the dust inside the collection groove 29 to prevent the dust from clogging the locking groove 15. When the locking block 14 slides outward toward the locking groove 15, the abutment of the locking block 14 against the drive block 28 gradually decreases until it disappears. During this process, the elastic force of the fifth spring 30 is released, thereby causing the drive block 28 to automatically reset. At the same time as the drive block 28 resets, it will generate a suction force inside the conduction channel 26, thereby causing the pusher 27 to reset synchronously.

[0080] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for 3D point cloud stitching based on a robot arm, characterized in that, S1: set the scanning path of the model to be calibrated; S2: acquire calibration pictures; S3: calibrate three-dimensional scanner parameters; S4: set the scanning path of the model reconstruction; S5: acquire point cloud data; S6: two-side point cloud splicing and fusion; S7: multi-view point cloud splicing and fusion; the mechanical arm is used for driving the three-dimensional scanner to realize scanning of the model with an internal hollow structure, comprising a mechanical arm body, the three-dimensional scanner is arranged on the execution end of the mechanical arm body, further comprising a fine adjustment mechanism for adjusting the rotation angle and axial movement between the three-dimensional scanner and the execution end of the mechanical arm body; The fine adjustment mechanism comprises a gas cylinder, an installation groove is formed in the mechanical arm body, the gas cylinder is arranged in the installation groove, a rotating seat is slidably arranged on the mechanical arm body, the rotating seat is connected with the output end of the gas cylinder, a rotating shaft is rotatably arranged on the rotating seat, and the three-dimensional scanner is fixedly connected with the rotating shaft; The fine adjustment mechanism further comprises a driving motor, the driving motor is slidably arranged on the rotating seat, the driving motor has a first position connected with the rotating shaft and a second position disconnected with the rotating shaft; further comprising a power assembly for driving the driving motor to switch between the first position and the second position; further comprising a locking assembly for fixing the rotating shaft when the driving motor is in the second position; The power assembly comprises a transmission frame, the transmission frame is fixedly connected with the base of the driving motor, a first wedge-shaped block is fixedly connected on the transmission frame, a second wedge-shaped block is fixedly connected on the mechanical arm body, the first wedge-shaped block is located in the movement stroke of the second wedge-shaped block, and a first spring is arranged between the transmission frame and the rotating seat; The locking assembly comprises a locking block, the locking block is arranged on the transmission frame, a locking groove matched with the locking block is formed in the rotating shaft, the locking block is away from the locking groove when the driving motor is in the first position, and the locking block is inserted into the locking groove when the driving motor is in the second position; A rotating hole is formed in the transmission frame, a limiting cylinder is fixedly connected on the inner side wall of the rotating hole, a limiting groove matched with the limiting cylinder is formed on the outer circumferential surface of the locking block, the limiting cylinder and the limiting groove are slidably connected, and a second spring is arranged between the locking block and the side wall of the rotating hole.

2. The method of claim 1, wherein, In step S1, based on the structure of two cameras and one projector, the model is placed in the specified placement area, the projector projects a beam of light onto the model, the position of the mechanical arm is adjusted based on the calibration board pictures taken by the two cameras, the positions of the three-dimensional scanner are determined, and the point positions of the mechanical arm at each calibration position are recorded for subsequent camera and projector calibration.

3. The method of claim 2, wherein, In step S2, first, the mechanical arm communicates with the three-dimensional scanner, after successful communication, the three-dimensional scanner is calibrated, then the mechanical arm moves according to the set calibration position, stops after moving to the specified position, waits for the three-dimensional scanner to take the calibration pictures of the current position, and then moves to the next calibration position, and returns to the zero position after all positions are completed.

4. The method of claim 3, wherein, In step S3, based on the calibration pictures taken by the three-dimensional scanner, the intrinsic matrix, distortion coefficient matrix and translation matrix of the camera and the projector are calibrated, and the rotation and translation relationship between the two cameras and between the two cameras and the projector is also calibrated, which is used for subsequent left and right point cloud splicing.

5. The method of claim 4, wherein, In step S4, the mechanical arm is controlled to move so that the three-dimensional scanner can realize all-around scanning of the model, and finally the relative position relationship between each scanning position is calibrated.

6. The method of claim 5, wherein, In step S5, after the mechanical arm successfully communicates with the three-dimensional scanner, the mechanical arm moves along the set scanning path. When it reaches a position, the three-dimensional scanner scans the model. After the scanning is completed, the mechanical arm moves to the next position. After all the positions are completed, the three-dimensional point cloud data of the model at each position is obtained.

7. The method of claim 6, wherein, In step S6, the two cameras and the projector form two monocular structured light systems respectively, and two point cloud data can be constructed at each point. The two point cloud data are obtained from two views of the model. Before the multi-view point cloud splicing, the two point cloud data corresponding to each point need to be fused to obtain a complete point cloud data.

8. The method of claim 7, wherein, In step S7, after the complete point cloud data under each view is obtained, according to the relative position relationship between each scanning position, the three-dimensional point cloud data under each view can be unified to the same view, so that a complete three-dimensional point cloud model is obtained.

Images