A multi-view three-dimensional point cloud imaging method and device for ceramic sanitary wares

Abstract

The present application relates to a kind of multi-view three-dimensional point cloud imaging method and device for ceramic sanitary ware, comprising the following: rotating platform keeps still, moving robot, obtains the point cloud information and transformation relationship of each view recorded by three-dimensional camera in multiple views under the surface information of ceramic sanitary ware blank perceived;Robot keeps still, rotates rotating platform, so that three-dimensional camera perceives the surface information of ceramic sanitary ware blank in multiple views, obtains the point cloud information and transformation relationship of each view;Repeat the above process until the complete surface information of ceramic sanitary ware is perceived;By machine vision processing system, the point cloud of misregistration is registered, and redundant point cloud is removed, i.e. complete point cloud model can be reconstructed.The present application can provide accurate spatial position information for subsequent ceramic sanitary ware production line operation, and lay a good foundation for the segmentation of polishing area and the planning of polishing path.

Description

Technical Field

[0001] This invention relates to the field of related technologies, and in particular to a multi-view three-dimensional point cloud imaging method and apparatus for ceramic sanitary ware. Background Technology

[0002] With the rapid development of automation technologies, robots are gradually replacing manual labor in production lines for ceramic sanitary ware grinding and spraying. Three-dimensional imaging equipment provides robots with identification and positioning information, making it an indispensable component of these production lines. Because ceramic sanitary ware is an unstructured object composed of multiple complex free-form surfaces, this unique structure leads to optical occlusion between different working surfaces. Furthermore, the limited field of view of imaging equipment makes it impossible to reconstruct the complete surface spatial information of ceramic sanitary ware from a single viewpoint when acquiring its 3D point cloud model. Therefore, it is necessary to perceive the surface of ceramic sanitary ware from multiple perspectives to obtain complete spatial information. However, the spatial information of the surface point cloud obtained by scanning ceramic sanitary ware from multiple perspectives using a 3D camera is misaligned, requiring precise registration to obtain a complete point cloud. However, registration algorithms are prone to getting trapped in local optima, resulting in significant deviations and making it difficult to obtain an accurate and complete point cloud. Currently, most multi-view imaging solutions utilize array cameras to acquire data, which involves cumbersome data processing and high computational complexity in data fusion. Therefore, it is necessary to develop a simple and efficient multi-view imaging method to simplify data acquisition and computation. Summary of the Invention

[0003] The purpose of this invention is to at least address one of the shortcomings of the prior art and provide a multi-view three-dimensional point cloud imaging method and device for ceramic sanitary ware.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] Specifically, a multi-view 3D point cloud imaging method for ceramic sanitary ware is proposed, including the following:

[0006] Using the robot base as the base coordinate system, a 3D camera is installed on the end flange of the robot, and the ceramic sanitary ware is placed on the rotating platform;

[0007] Step 110: Keep the rotating platform stationary, move the robot, and acquire the point cloud information and transformation relationship of each viewpoint recorded by the 3D camera from multiple perspectives of the surface information of the ceramic sanitary ware blank.

[0008] Step 120: The robot remains stationary while rotating the rotating platform, allowing the 3D camera to perceive the surface information of the ceramic sanitary ware blank from multiple perspectives and obtain the point cloud information and transformation relationship from each perspective.

[0009] Step 130: Repeat steps 110 to 120 until the complete surface information of the ceramic sanitary ware is perceived;

[0010] Step 140: Perform multi-view point cloud coarse registration on the complete surface information of the ceramic sanitary ware to obtain the coarse registration point cloud;

[0011] Step 150: Perform multi-view fine registration of the coarsely registered point cloud to obtain a finely registered point cloud;

[0012] Step 160: Perform point cloud processing on the precise point cloud to obtain a complete model.

[0013] Furthermore, specifically, coarse registration of the complete surface information of ceramic sanitary ware is performed using multi-view point clouds to obtain a coarsely registered point cloud, including:

[0014] With a fixed viewpoint, an object is placed on a rotating platform. An imaging device can acquire a point on the object's surface in the camera coordinate system {C}. If the rotation angle is... Let the current point be... , represents a column vector, where points are represented as the rotating platform rotates. The positions are always distributed on a spatial circle. Taking the center O of the current spatial circle as the center of rotation, the unit normal vector of the plane containing the spatial circle is... As the direction vector of the rotation axis, let the equation of the plane of the current spatial circle under {C} be: Then we have:

[0015] ;

[0016] Substituting a series of observation points into formula (5) yields the result. After normalizing v, we can obtain the result. If a chord on a spatial circle is perpendicularly bisected by a line perpendicular from the center of the circle to the chord, then: ;

[0017] Turning points of a series of observation points Substituting into formulas (5) and (6), the center of rotation O can be obtained; let a point be located under {B}. Rotate around the center of rotation O and the axis of rotation We obtain a new point that can be transformed into a point rotated about the origin and the axis of rotation in {W}. This problem can be described using the Rodriguez rotation formula, and thus solved. In ,pass , , This allows for obtaining a good initial pose of the point cloud from multiple perspectives, enabling coarse registration of the point cloud model under arbitrary robot poses and rotating platform angles.

[0018] Furthermore, specifically, the conversion process of formula (2) includes,

[0019] The robot's base coordinate system is {B}, the robot's end effector coordinate system is {E}, the imaging device's modeling coordinate system is {C}, the rotating platform's zero-point coordinate system is {R}, and the world coordinate system is {W}. A point based on the robot's base coordinate system... If it can be observed from different perspectives, then:

[0020] ;

[0021] In formula (1), It is the point in the modeling coordinate system of the imaging device at the current viewpoint. coordinates It is the homogeneous transformation matrix from the robot's base coordinate system to the end flange. Let {E} be the homogeneous transformation matrix from coordinate system {C}. When the robot's posture remains constant and the rotating platform rotates by β, to correct the imaging under different rotational viewpoints, it is necessary to... Rotate β in the opposite direction of the rotation axis of the rotating platform so that it is located in the zero coordinate system {R}; when the robot is in an arbitrary posture and the rotating platform is in an arbitrary angle, if the world coordinate system {W} coincides with the zero coordinate system {R} of the rotating platform, the arbitrary scanning point can be obtained by combining formula (1). Representation in the world coordinate system:

[0022] ;

[0023] In formula (2), This represents the homogeneous transformation matrix for rotating β in the opposite direction around the rotation axis of the rotating platform. Through forward kinematics analysis of the robot, the DH parameters are established based on the robot's model parameters, and the transformations between its joints are calculated. This yields the transformation matrix from the robot's base coordinate system to the end flange. .

[0024] Furthermore, specifically, this invention employs an "eye-on-hand" method to install the imaging device, and the transformation relationship from the robot's end-effector coordinate system {E} to the imaging device's coordinate system {C} is referred to as the hand-eye matrix. If the rotating platform does not rotate, but only the robot moves, and the point clouds imaged from viewpoints a and b are unified to the base coordinate system {B}, and there is an overlap between the two point cloud clusters, then the overlap satisfies the following formula:

[0025] ;

[0026] Since the imaging device is fixed to the end flange of the robot, and They are the same transformation matrix, denoted as .Will express{ }arrive{ If the transformation matrix of} is given, then formula (3) can be expressed as:

[0027] ;

[0028] The transformation matrix from coordinate system {B} to coordinate system {E} is obtained through the robot's forward kinematics. In formula (4), and The answer can be obtained by substituting the robot joint angles from the corresponding viewpoints into the robot transformation matrix. At this point, only the following needs to be determined: The hand-eye matrix can then be obtained. Solving rigid body transformations The problem is transformed into solving the matching problem of point clouds in different coordinate systems; the Iterative Closest Point (ICP) algorithm is used to calculate the transformation matrix of the two point cloud clusters, thereby obtaining... Substituting it into formula (4) yields the hand-eye matrix. .

[0029] Furthermore, specifically, the coarsely registered point cloud is subjected to multi-view point cloud fine registration to obtain a finely registered point cloud, including:

[0030] For each point in the source point cloud P Find the nearest neighbor in the target point cloud Q. The square of its Euclidean distance is calculated as follows:

[0031] ;

[0032] Let the number of overlapping point clouds be The number of source point clouds is ,Will Sort in ascending order and extract the first few bytes. The mean square error (MSE) is calculated at each point during the iterative process; the transformation matrix is ​​continuously iterated and updated until the MSE is less than a certain threshold, at which point the algorithm converges and the accurate rotation matrix R and translation vector t can be obtained.

[0033] ;

[0034] The multi-view point cloud fine registration process includes the following:

[0035] For point clouds at different viewpoints under the same robot pose, registration is performed. Using a rotation angle β=0 as the reference, point clouds at other rotation angles are registered to this reference. If the rotation angle... Register sequentially with the reference viewpoint; if the angle is... Register backwards to the reference viewpoint in sequence;

[0036] Point clouds at all rotation angles β=0 under different robot poses are registered. A point cloud at rotation angle β=0 in a robot state is selected as the reference, and point clouds at different robot poses are registered to this reference. The rotation matrix of each registration is recorded. Translation vector ;

[0037] Apply the corresponding rotation matrix to the point cloud under the same robot state except for the view with β=0. Translation vector This allows for accurate registration of global multi-view point clouds.

[0038] This invention also proposes a multi-view three-dimensional point cloud imaging device for ceramic sanitary ware, comprising the following:

[0039] Using the robot base as the base coordinate system, a 3D camera is installed on the end flange of the robot, and the ceramic sanitary ware is placed on a rotating platform at a preset position;

[0040] The first data acquisition module is used to acquire point cloud information and transformation relationships from each perspective of the 3D camera, which perceives the surface information of the ceramic sanitary ware blank from multiple perspectives while the rotating platform remains stationary and the robot is moving.

[0041] The second data acquisition module is used to keep the robot stationary and rotate the rotating platform so that the 3D camera can perceive the surface information of the ceramic sanitary ware blank from multiple perspectives and acquire the point cloud information and transformation relationship from each perspective.

[0042] Repeatedly run the first data acquisition module and the second data acquisition module until complete surface information of the ceramic sanitary ware is perceived;

[0043] The coarse registration module is used to perform multi-view point cloud coarse registration on the complete surface information of ceramic sanitary ware to obtain a coarse registration point cloud;

[0044] The fine registration module is used to perform multi-view fine registration of the coarse registration point cloud to obtain a fine registration point cloud;

[0045] The complete model acquisition module is used to perform point cloud processing on the precise point cloud to obtain a complete model.

[0046] The beneficial effects of this invention are as follows:

[0047] This invention proposes a multi-view 3D point cloud imaging method and device for ceramic sanitary ware. By fixing a 3D camera to the end flange of a robot and coordinating with the rotation of a rotating platform, the surface information of the ceramic sanitary ware can be perceived from all angles, solving the problem that a single viewpoint cannot perceive the complete surface information of the ceramic sanitary ware. Furthermore, this invention utilizes only a single imaging device to achieve multi-view point cloud imaging, reducing the need for additional imaging equipment and simplifying the system hardware structure. Moreover, by using automated equipment such as robots and rotating platforms for point cloud data acquisition, complex manual operations are avoided, simplifying the process flow and improving work efficiency and quality.

[0048] The method of this patent can quickly and automatically reconstruct a complete and accurate point cloud model, providing precise spatial location information for subsequent operations on the ceramic sanitary ware production line, and laying a good foundation for the segmentation of the grinding area and the planning of the grinding path. Attached Figure Description

[0049] The above and other features of this disclosure will become more apparent from the detailed description of the embodiments illustrated in conjunction with the accompanying drawings. In the accompanying drawings, the same reference numerals denote the same or similar elements. Obviously, the drawings described below are merely some embodiments of this disclosure. Those skilled in the art can obtain other drawings based on these drawings without any creative effort. In the drawings:

[0050] Figure 1 The diagram shown is a flowchart of a multi-view three-dimensional point cloud imaging method for ceramic sanitary ware according to the present invention.

[0051] Figure 2 (a) shows a schematic diagram of imaging a ceramic sanitary ware blank from a single perspective in one embodiment, and (b) shows a schematic diagram of imaging a ceramic sanitary ware blank from multiple perspectives in one embodiment.

[0052] Figure 3 The diagram shown illustrates the working principle of a multi-view 3D point cloud imaging device.

[0053] Figure 4 The diagram shown is a flowchart for solving the coordinate transformation relationship;

[0054] Figure 5 The diagram shows the principle of transformation between two different 3D imaging coordinate systems.

[0055] Figure 6 The diagram shown is a schematic of the rotation of an object on a rotating platform.

[0056] Figure 7 The diagram shown illustrates a multi-view point cloud registration method.

[0057] Figure 8(a) is a point cloud of a ceramic sanitary ware blank in one embodiment, and (b) is a point cloud of a ceramic sanitary ware blank in another embodiment. Detailed Implementation

[0058] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The same reference numerals used throughout the accompanying drawings indicate the same or similar parts.

[0059] Example 1, referring to Figure 1 This invention proposes a multi-view 3D point cloud imaging method for ceramic sanitary ware, including the following:

[0060] Using the robot base as the base coordinate system, a 3D camera is installed on the end flange of the robot, and the ceramic sanitary ware is placed on the rotating platform;

[0061] Step 110: Keep the rotating platform stationary, move the robot, and acquire the point cloud information and transformation relationship of each viewpoint recorded by the 3D camera from multiple perspectives of the surface information of the ceramic sanitary ware blank.

[0062] Step 120: The robot remains stationary while rotating the rotating platform, allowing the 3D camera to perceive the surface information of the ceramic sanitary ware blank from multiple perspectives and obtain the point cloud information and transformation relationship from each perspective.

[0063] Step 130: Repeat steps 110 to 120 until the complete surface information of the ceramic sanitary ware is perceived;

[0064] Step 140: Perform multi-view point cloud coarse registration on the complete surface information of the ceramic sanitary ware to obtain the coarse registration point cloud;

[0065] Step 150: Perform multi-view fine registration of the coarsely registered point cloud to obtain a finely registered point cloud;

[0066] Step 160: Perform point cloud processing on the precise point cloud to obtain a complete model.

[0067] In this embodiment 1, by fixing a 3D camera to the end flange of the robot and coordinating with the rotation of the rotating platform, the surface information of the ceramic sanitary ware can be perceived from all angles, solving the problem that the complete surface information of the ceramic sanitary ware cannot be perceived from a single viewpoint. Furthermore, this invention can achieve multi-view point cloud imaging using only a single imaging device, reducing the need for imaging equipment and simplifying the system hardware structure. Moreover, by using automated equipment such as robots and rotating platforms for point cloud data acquisition, complex manual operations are avoided, simplifying the process flow and improving work efficiency and quality.

[0068] The method of this patent can quickly and automatically reconstruct a complete and accurate point cloud model, providing precise spatial location information for subsequent operations on the ceramic sanitary ware production line, and laying a good foundation for the segmentation of the grinding area and the planning of the grinding path.

[0069] The implementation process of this invention mainly includes the following:

[0070] 1. Analysis and Construction of 3D Imaging System

[0071] Ceramic sanitary ware is an unstructured object composed of multiple complex curved surfaces. Due to its inherent structural characteristics, it exhibits optical self-occlusion, and because of the limited field of view of imaging equipment, it is impossible to fully perceive all information from a single perspective. Figure 2 As shown in (a). To perceive complete surface information of the unglazed ceramic sanitary ware, one can do so as shown in (a). Figure 2 As shown in (b), complete information is obtained through perception from multiple perspectives, and then the three-dimensional point cloud model is reconstructed based on the relative spatial information of each perspective.

[0072] This invention fixes the imaging device to the end flange of a robot and, in conjunction with a rotating platform, enables the imaging device to perceive complete model information, such as... Figure 3 As shown. The robot's base coordinate system is {B}, the robot's end effector coordinate system is {E}, the imaging device's modeling coordinate system is {C}, the rotating platform's zero-point coordinate system is {R}, and the world coordinate system is {W}. Based on a point in the robot's base coordinate system... If it can be observed from different perspectives, then:

[0073] ;

[0074] In formula (1), It is the point in the modeling coordinate system of the imaging device at the current viewpoint. coordinates It is the homogeneous transformation matrix from the robot's base coordinate system to the end flange. Let {E} be the homogeneous transformation matrix from coordinate system {C}. When the robot's posture remains constant and the rotating platform rotates by β, to correct the imaging under different rotational viewpoints, it is necessary to... Rotate β in the opposite direction of the rotation axis of the rotating platform until it is located in the zero coordinate system {R}. When the robot is in an arbitrary posture and the rotating platform is at an arbitrary angle, if we assume that the world coordinate system {W} coincides with the zero coordinate system {R} of the rotating platform, we can obtain any scanning point by combining formula (1). Representation in the world coordinate system:

[0075] ;

[0076] In formula (2), This represents the homogeneous transformation matrix of rotating β in the opposite direction around the rotation axis of the rotating platform. Using the above formula, the point cloud obtained from any single scan can be transformed into the world coordinate system {W}, achieving coarse registration of the point cloud and obtaining a three-dimensional point cloud model of the ceramic sanitary ware blank. Regarding the spatial transformation relationship matrix in formulas (1) and (2)... , and According to Figure 4 The flowcharts shown are used to perform the calculations.

[0077] 2. Forward Kinematics of Robots

[0078] By analyzing the robot's forward kinematics, establishing the DH parameters based on the robot's model parameters, and calculating the transformations between its joints, the transformation matrix from the robot's base coordinate system to the end flange can be obtained. .

[0079] 3. Hand-eye calibration method based on point cloud registration

[0080] like Figure 3 As shown, this invention uses an "eye-on-hand" method to mount the imaging device. The transformation relationship from the robot's end-effector coordinate system {E} to the imaging device's coordinate system {C} is called the hand-eye matrix. If the rotating platform remains stationary, and only the robot moves, unifying the point clouds imaged from viewpoints a and b to the base coordinate system {B}, and there is an overlap between the two point cloud clusters, then the overlap satisfies the following formula:

[0081] ;

[0082] like Figure 3 As shown, the imaging device is fixed to the end flange of the robot, therefore and They are the same transformation matrix, denoted as .Will express{ }arrive{ If the transformation matrix of} is given, then formula (3) can be expressed as:

[0083] ;

[0084] The transformation matrix from coordinate system {B} to coordinate system {E} is obtained through the robot's forward kinematics. In formula (4), and The answer can be obtained by substituting the robot joint angles from the corresponding viewpoints into the robot transformation matrix. At this point, only the following needs to be determined: The hand-eye matrix can then be obtained. .like Figure 5 As shown, the rigid body transformation can be solved. The problem is transformed into solving the matching problem of point clouds in different coordinate systems. This invention uses the iterative closest point (ICP) algorithm to calculate the transformation matrix of the two point cloud clusters, thereby obtaining... Substituting it into formula (4) yields the hand-eye matrix. .

[0085] 4. Rotary platform calibration method

[0086] like Figure 6 As shown, with a fixed viewing angle, an object is placed on a rotating platform. An imaging device can acquire a point on the object's surface in the camera coordinate system {C}. If the rotation angle is... Let the current point be... , represents a column vector, where points are represented as the rotating platform rotates. The positions are always distributed on a spatial circle. Taking the center O of the current spatial circle as the center of rotation, the unit normal vector of the plane containing the spatial circle is... As the direction vector of the rotation axis, let the equation of the plane of the current spatial circle under {C} be: Then we have:

[0087] ;

[0088] Substituting a series of observation points into formula (5) yields the result. After normalizing v, we can obtain the result. If a chord on a spatial circle is perpendicularly bisected by a line perpendicular from the center of the circle to the chord, then: ;

[0089] Turning points of a series of observation points Substituting into formulas (5) and (6), the center of rotation O can be obtained. Let a point be located under {B}. Rotate around the center of rotation O and the axis of rotation We obtain a new point that can be transformed into a point rotated about the origin and the axis of rotation under {W}. This problem can be described using the Rodriguez rotation formula, which yields the value in formula (2). .pass , , This allows for obtaining a good initial pose of the point cloud from multiple perspectives, enabling coarse registration of the point cloud model under any robot posture and any angle of the rotating platform.

[0090] 5. Multi-view point cloud fine registration method

[0091] The above steps have achieved coarse registration of point clouds from multiple perspectives. However, since the overlapping areas of point clouds from different perspectives are small, the clusters of point clouds in the non-overlapping areas can significantly affect the accuracy of the registration results. This invention uses the TrICP algorithm to solve this problem. For each point in the source point cloud P... Find the nearest neighbor in the target point cloud Q. The square of its Euclidean distance is calculated as follows:

[0092] ;

[0093] Let the number of overlapping point clouds be The number of source point clouds is ,Will Sort in ascending order and extract the first few bytes. The mean squared error (MSE) is calculated at each point during the iterative process. The transformation matrix is ​​updated continuously until the MSE is less than a certain threshold, at which point the algorithm converges. This yields the accurate rotation matrix R and translation vector t.

[0094] ;

[0095] The precise registration process of multi-view point clouds is as follows Figure 7 As shown, it mainly includes three steps:

[0096] Register point clouds from different viewpoints within the same robot pose. Using a rotation angle β=0 as a reference, register point clouds from other rotation angles to this reference. If the rotation angle... Register sequentially with the reference viewpoint; if the angle is... Then, register backwards to the reference viewpoint.

[0097] Point clouds at all rotation angles β=0 under different robot poses are registered. A point cloud at rotation angle β=0 in a given robot state is selected as the reference, and point clouds from different robot poses are registered to this reference. The rotation matrix for each registration is recorded. Translation vector .

[0098] Apply the corresponding rotation matrix to the point cloud under the same robot state except for the view with β=0. Translation vector This allows for accurate registration of global multi-view point clouds.

[0099] 6. Experimental Results

[0100] An imaging device is fixed at the end effector of a robot, and a rotating platform is used to perceive the complete surface information of the ceramic sanitary ware blank. The point cloud information from various viewpoints is fused using the method of this invention, and redundant information is removed. After these processes, a complete surface image can be reconstructed. Figure 8The complete point cloud model of the unfinished ceramic sanitary ware shown in (a) and (b) is shown.

[0101] It should be noted that the capital letter subscript under any English letter refers to its meaning within the coordinate system represented by that capital letter, and and They represent in The transformation matrix from the robot's base coordinate system to the end flange from the viewpoint. and They respectively represent their in Combination of robot joint angles from a different perspective.

[0102] like Figure 5 As shown, { }and{ } represent the imaging coordinate systems from viewpoints a and b, respectively. and These represent the imaging points at { }and{ Coordinates in the coordinate system.

[0103] This invention also proposes a multi-view three-dimensional point cloud imaging device for ceramic sanitary ware, comprising the following:

[0104] Using the robot base as the base coordinate system, a 3D camera is installed on the end flange of the robot, and the ceramic sanitary ware is placed on a rotating platform at a preset position;

[0105] The first data acquisition module is used to acquire point cloud information and transformation relationships from each perspective of the 3D camera, which perceives the surface information of the ceramic sanitary ware blank from multiple perspectives while the rotating platform remains stationary and the robot is moving.

[0106] The second data acquisition module is used to keep the robot stationary and rotate the rotating platform so that the 3D camera can perceive the surface information of the ceramic sanitary ware blank from multiple perspectives and acquire the point cloud information and transformation relationship from each perspective.

[0107] Repeatedly run the first data acquisition module and the second data acquisition module until complete surface information of the ceramic sanitary ware is perceived;

[0108] The coarse registration module is used to perform multi-view point cloud coarse registration on the complete surface information of ceramic sanitary ware to obtain a coarse registration point cloud;

[0109] The fine registration module is used to perform multi-view fine registration of the coarse registration point cloud to obtain a fine registration point cloud;

[0110] The complete model acquisition module is used to perform point cloud processing on the precise point cloud to obtain a complete model.

[0111] The basic components of this invention include: a 3D camera, a six-axis robot, a rotating platform, and a machine vision processing system.

[0112] A 3D camera is used for 3D point cloud imaging from the current viewpoint;

[0113] The six-axis robot is responsible for moving the 3D camera fixed to the end of the robot to the appropriate imaging position;

[0114] The rotating platform is used to rotate the unfinished ceramic sanitary ware so that it can be completely perceived by the 3D camera;

[0115] The machine vision processing system processes the raw data from the 3D camera to reconstruct a complete point cloud model.

[0116] When applying,

[0117] (a) Using the robot base as the base coordinate system, install a 3D camera on the end flange of the robot and place the ceramic sanitary ware on the rotating platform;

[0118] (ii) The rotating platform remains stationary while the robot moves, allowing the 3D camera to perceive the surface information of the ceramic sanitary ware blank from multiple perspectives, and record the point cloud information and transformation relationship from the current perspective;

[0119] (iii) The robot remains stationary while rotating the rotating platform, allowing the 3D camera to perceive a wider range of ceramic sanitary ware surface information, record point cloud information and transformation relationships at different rotation angles;

[0120] (iv) Steps (ii) and (iii) are performed alternately until the complete surface information of the ceramic sanitary ware is perceived. The misaligned point cloud is registered by the machine vision processing system and redundant point cloud is removed, so that the complete point cloud model can be reconstructed.

[0121] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0122] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0123] Although the description of the invention has been quite detailed and particularly of several described embodiments, it is not intended to limit it to any of these details or embodiments or any particular embodiment, but should be considered as providing a broad possible interpretation of the claims by referring to the appended claims and taking into account the prior art, thereby effectively covering the intended scope of the invention. Furthermore, the invention has been described above with respect to embodiments foreseeable by the inventors in order to provide a useful description, and non-substantial modifications to the invention that have not yet been foreseen may still represent equivalent modifications.

[0124] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. Any embodiment that achieves the technical effects of the present invention using the same means should fall within the protection scope of the present invention. Within the protection scope of the present invention, various modifications and variations of the technical solutions and / or embodiments are possible.

Claims

1. A multi-view 3D point cloud imaging method for ceramic sanitary ware, characterized in that, Including the following: Using the robot base as the base coordinate system, a 3D camera is installed on the end flange of the robot, and the ceramic sanitary ware is placed on the rotating platform; Step 110: Keep the rotating platform stationary, move the robot, and acquire the point cloud information and transformation relationship of each viewpoint recorded by the 3D camera from multiple perspectives of the surface information of the ceramic sanitary ware blank. Step 120: The robot remains stationary while rotating the rotating platform, allowing the 3D camera to perceive the surface information of the ceramic sanitary ware blank from multiple perspectives and obtain the point cloud information and transformation relationship from each perspective. Step 130: Repeat steps 110 to 120 until the complete surface information of the ceramic sanitary ware is perceived; Step 140: Perform multi-view point cloud coarse registration on the complete surface information of the ceramic sanitary ware to obtain the coarse registration point cloud; Step 150: Perform multi-view fine registration of the coarsely registered point cloud to obtain a finely registered point cloud; Step 160: Perform point cloud processing on the precisely matched point cloud to obtain a complete model; Specifically, coarse registration point clouds are obtained by performing multi-view point cloud coarse registration on the complete surface information of ceramic sanitary ware. include, With a fixed viewpoint, an object is placed on a rotating platform. An imaging device can acquire a point on the object's surface in the camera coordinate system {C}. If the rotation angle is... Let the current point be... , represents a column vector, where points are shifted as the rotating platform rotates. The positions are always distributed on a spatial circle, with the center O of the current spatial circle as the center of rotation, and the unit normal vector of the plane containing the spatial circle. As the direction vector of the rotation axis, let the equation of the plane of the current spatial circle under {C} be: Then we have: ； Substituting a series of observation points into formula (5) yields the result. After normalizing v, we can obtain the result. If a chord on a spatial circle is perpendicularly bisected by a line perpendicular from the center of the circle to the chord, then: ; Turning points of a series of observation points Substituting into formulas (5) and (6) yields the center of rotation O; Let a point be located under {B} Rotate around the center of rotation O and the axis of rotation We obtain a new point that can be transformed into a point rotated about the origin and the axis of rotation in {W}. This can be described using the Rodriguez rotation formula, i.e., the result is obtained. In ,pass , , This allows us to obtain the initial pose of the point cloud from multiple perspectives, enabling coarse registration of the point cloud model under arbitrary robot poses and rotating platform angles. Here, {B} and {W} represent the robot's base coordinate system and world coordinate system, respectively. This represents the homogeneous transformation matrix that rotates the platform in the opposite direction of the rotation axis by β. It is the homogeneous transformation matrix from the robot's base coordinate system to the end flange. Let be the homogeneous transformation matrix from coordinate system {E} to {C}. It is the point of the imaging device in the camera coordinate system at the current viewpoint. The coordinates.

2. The multi-view three-dimensional point cloud imaging method for ceramic sanitary ware according to claim 1, characterized in that, Specifically, the conversion process of formula (2) includes, The robot's base coordinate system is {B}, the robot's end effector coordinate system is {E}, the imaging device's camera coordinate system is {C}, the rotating platform's zero-point coordinate system is {R}, and the world coordinate system is {W}. A point based on the robot's base coordinate system... If it can be observed from different perspectives, then: ； In formula (1), It is the point of the imaging device in the camera coordinate system at the current viewpoint. coordinates It is the homogeneous transformation matrix from the robot's base coordinate system to the end flange. Let {E} be the homogeneous transformation matrix from coordinate system {C}. When the robot's posture remains constant and the rotating platform rotates by β, to correct the imaging under different rotational viewpoints, it is necessary to... Rotate β in the opposite direction of the rotation axis of the rotating platform so that it is located in the zero coordinate system {R}; when the robot is in an arbitrary posture and the rotating platform is in an arbitrary angle, if the world coordinate system {W} coincides with the zero coordinate system {R} of the rotating platform, the arbitrary scanning point can be obtained by combining formula (1). Representation in the world coordinate system: ； In formula (2), This represents the homogeneous transformation matrix for rotating β in the opposite direction around the rotation axis of the rotating platform. Through forward kinematics analysis of the robot, the DH parameters are established based on the robot's model parameters, and the transformations between its joints are calculated. This yields the transformation matrix from the robot's base coordinate system to the end flange. .

3. The multi-view three-dimensional point cloud imaging method for ceramic sanitary ware according to claim 2, characterized in that, Specifically, the imaging device is installed using the "eye on hand" method. The transformation relationship between the robot's end effector coordinate system {E} and the imaging device's coordinate system {C} is called the hand-eye matrix. If the rotating platform does not rotate, but only the robot moves, and the point clouds imaged from viewpoints a and b are unified to the base coordinate system {B}, and there is an overlap between the two point cloud clusters, then the overlap satisfies the following formula: ； Since the imaging device is fixed to the end flange of the robot, and They are the same transformation matrix, denoted as ,Will express{ }arrive{ If the transformation matrix of} is given, then formula (3) can be expressed as: ； The transformation matrix from coordinate system {B} to coordinate system {E} is obtained through the robot's forward kinematics. In formula (4), and The answer can be obtained by substituting the robot joint angles from the corresponding viewpoints into the robot transformation matrix. At this point, only the following needs to be determined: The hand-eye matrix can then be obtained. Solving rigid body transformations The problem is transformed into solving the matching problem of point clouds in different coordinate systems; The Iterative Closest Point (ICP) algorithm is used to calculate the transformation matrix of the two point cloud clusters, thereby obtaining... Substituting it into formula (4) yields the hand-eye matrix. .

4. The multi-view three-dimensional point cloud imaging method for ceramic sanitary ware according to claim 3, characterized in that, Specifically, the coarsely registered point cloud is subjected to multi-view point cloud fine registration to obtain a finely registered point cloud. include, For each point in the source point cloud P Find the nearest neighbor in the target point cloud Q. The square of its Euclidean distance is calculated as follows: ； Let the number of overlapping point clouds be The number of source point clouds is ,Will Sort in ascending order and extract the first few bytes. The mean square error (MSE) is calculated at each point during the iterative process. By iterating and updating the transformation matrix until the MSE is less than a certain threshold, the algorithm converges and the accurate rotation matrix R and translation vector t can be obtained. ； The multi-view point cloud fine registration process includes the following: For point clouds at different viewpoints under the same robot pose, registration is performed. Using a rotation angle β=0 as the reference, point clouds at other rotation angles are registered to this reference. If the rotation angle... Register sequentially with the reference viewpoint; if the angle is... Register backwards to the reference viewpoint in sequence; Point clouds at all rotation angles β=0 under different robot poses are registered. A point cloud at rotation angle β=0 in a robot state is selected as the reference, and point clouds at different robot poses are registered to this reference. The rotation matrix of each registration is recorded. Translation vector ; Apply the corresponding rotation matrix to the point cloud under the same robot state except for the view with β=0. Translation vector This allows for accurate registration of global multi-view point clouds.

5. A multi-view three-dimensional point cloud imaging device for ceramic sanitary ware, characterized in that, The apparatus comprising the steps of the method according to any one of claims 1-4, wherein the method is applied, and the apparatus includes the following: Using the robot base as the base coordinate system, a 3D camera is installed on the end flange of the robot, and the ceramic sanitary ware is placed on a rotating platform at a preset position; The first data acquisition module is used to acquire point cloud information and transformation relationships from each perspective of the 3D camera, which perceives the surface information of the ceramic sanitary ware blank from multiple perspectives while the rotating platform remains stationary and the robot is moving. The second data acquisition module is used to keep the robot stationary and rotate the rotating platform so that the 3D camera can perceive the surface information of the ceramic sanitary ware blank from multiple perspectives and acquire the point cloud information and transformation relationship from each perspective. Repeatedly run the first data acquisition module and the second data acquisition module until complete surface information of the ceramic sanitary ware is perceived; The coarse registration module is used to perform multi-view point cloud coarse registration on the complete surface information of ceramic sanitary ware to obtain a coarse registration point cloud; The fine registration module is used to perform multi-view fine registration of the coarse registration point cloud to obtain a fine registration point cloud; The complete model acquisition module is used to perform point cloud processing on the precise point cloud to obtain a complete model.

Images