Efficient 3D visual positioning method and device for large casting grinding

Abstract

The present invention relates to an efficient three-dimensional visual positioning method for large casting polishing, comprising the following steps: completing the calibration of a 3D industrial camera and a rotating platform so that the 3D industrial camera is aligned with a target casting; performing multi-perspective scanning of the target casting according to a preset single rotation angle and rotation direction, and obtaining casting point cloud data of multiple different perspectives according to the number of shots taken when the rotating platform rotates 360 degrees; performing multi-perspective point cloud registration, registering the obtained casting point cloud data of multiple different perspectives to the robot world coordinate system, and obtaining a globally reconstructed casting point cloud model; and completing local information missing on the globally reconstructed casting point cloud model to complete visual positioning. The present invention utilizes a machine vision processing system to perform point cloud registration and complete the reconstruction of the casting three-dimensional point cloud model, thereby achieving the spatial position required for precise positioning processing by the polishing robot. The method has high positioning accuracy and fast positioning speed.

Description

Technical Field

[0001] The present invention relates to the technical field of robot visual positioning, and in particular to a high-efficiency three-dimensional visual positioning method and device for large casting grinding. Background Art

[0002] While the number of foundries in my country continues to grow and casting production continues to surge, many still have relatively outdated casting equipment. Large castings are prone to surface defects such as flash, burrs, and pouring risers during production, necessitating post-processing techniques such as cutting, milling, and grinding to further improve surface quality. Large castings often require positioning before grinding.

[0003] Most companies use industrial robots to teach and reproduce castings or reverse engineer casting models using offline programming to polish castings. These traditional methods lack flexibility and are time-consuming, making them inadequate for practical, step-by-step polishing. Using 3D vision to reconstruct casting models to guide robotic polishing promotes green production, avoids hazards, improves efficiency, and addresses the shortcomings of the previous two methods. Summary of the Invention

[0004] The purpose of the present invention is to solve at least one of the deficiencies of the prior art and to provide an efficient three-dimensional visual positioning method and device for grinding large castings.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] Specifically, an efficient 3D visual positioning method for large casting grinding is proposed, including the following:

[0007] Complete the calibration of the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting;

[0008] Performing multi-view scanning on the target casting according to a preset single rotation angle and rotation direction, and obtaining casting point cloud data of multiple different view angles according to the number of shots when the rotating platform rotates 360 degrees;

[0009] Perform multi-view point cloud registration, register the casting point cloud data obtained from multiple different viewpoints to the robot's world coordinate system, and obtain a globally reconstructed casting point cloud model;

[0010] The globally reconstructed casting point cloud model is used to complete local information missing and complete visual positioning.

[0011] Furthermore, specifically, the calibration of the 3D industrial camera and the rotating platform is completed in the following way:

[0012] A visual standard plate is placed on the spindle center of the rotating platform, and the 3D industrial camera is focused so as to align the 3D industrial camera with the target casting.

[0013] Furthermore, specifically, the preset single rotation angle is 60 degrees, the rotation direction is forward, and the corresponding number of shots is 6 times.

[0014] Furthermore, specifically, the multi-view point cloud registration process includes the following:

[0015] Calculate the overlap W of any two viewpoint clouds. Assume that the number of 3D points in the two viewpoint clouds is m and n respectively. Use KD tree to detect all similar points in the two viewpoint clouds. Assume that the number of similar points is N. Then the overlap W of the two viewpoint clouds is calculated as follows:

[0016] The two viewpoint clouds that meet the overlap requirement W>0.2 are registered. Assume that the two viewpoint clouds that meet the overlap requirement are P={p1,...,p n}, Q={q1,...,q n}, now we want to find a Euclidean transformation R,t such that Constructing the least squares problem When the objective function reaches the minimum, the optimal rotation matrix R and translation matrix t are solved.

[0017] Defining Error Set the maximum number of iterations to 50. When the number of iterations is greater than 50 or e < 0.001, stop the iteration. Otherwise, enter the iteration.

[0018] Output the true value of the rotation and translation matrix and draw the registered image to obtain the globally reconstructed casting point cloud model.

[0019] Furthermore, specifically, the globally reconstructed casting point cloud model is supplemented with local information loss, including the following: the local surface of the target casting is perceived twice from different perspectives by an industrial camera to obtain a local dense point cloud model.

[0020] The global model is registered with the local dense model to build a weighted ICP point cloud registration model: Assume that there is P'={p'1,p'2,...,p' n} and Q'={q'1,q'2,...,q' n}Two sets of corresponding points, want to find a Euclidean transformation R,t,

[0021] So that: where ω i >0 is the weight corresponding to each point pair,

[0022] The mathematical model of solving weights is combined with the Welsh function, where the mathematical relationship between the Welsh function and weights is as follows: As for the value of coefficient ν, since there are more registration points and the distance between the points is larger in the initial registration stage, ν = 2 with a smoother curve is selected as the coefficient of the first registration. After each iteration, the proportional coefficient changes to 0.9 times the previous one, that is, ν n =0.9ν n-1 ,

[0023] According to the above, iteratively remove outliers and complete the missing local information.

[0024] The present invention also proposes an efficient three-dimensional visual positioning device for large casting grinding, comprising:

[0025] A calibration module is used to calibrate the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting;

[0026] A point cloud data acquisition module is used to perform multi-view scanning on the target casting according to a preset single rotation angle and rotation direction, and obtain casting point cloud data of multiple different perspectives according to the number of shots taken when the rotating platform rotates 360 degrees;

[0027] The point cloud registration module is used to perform multi-view point cloud registration, registering the casting point cloud data obtained from multiple different viewpoints to the robot's world coordinate system to obtain a globally reconstructed casting point cloud model;

[0028] The missing information completion module is used to complete local information missing in the globally reconstructed casting point cloud model and complete visual positioning.

[0029] The present invention also proposes an efficient three-dimensional vision platform for large casting grinding, which applies any of the above-mentioned efficient three-dimensional vision positioning methods for large casting grinding, including:

[0030] The robot body is used to fix the camera at the end and adjust the camera's shooting height and angle, so that the robot can process the casting after positioning;

[0031] A 3D industrial camera is provided at one end of the robot body and is used to capture point cloud data of castings from multiple different perspectives;

[0032] A rotating platform, adapted to the robot body and disposed below the 3D industrial camera, is used to place a target casting and drive the target casting to rotate;

[0033] The host computer is used to complete the calibration work of the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting.

[0034] The target casting is scanned from multiple perspectives according to the preset single rotation angle and rotation direction, and the casting point cloud data of multiple different perspectives is obtained according to the number of shots when the rotating platform rotates 360 degrees.

[0035] Perform multi-view point cloud registration, register the casting point cloud data obtained from multiple different viewpoints to the robot world coordinate system, and obtain a globally reconstructed casting point cloud model.

[0036] The globally reconstructed casting point cloud model is used to complete local information missing and visual positioning;

[0037] The host computer is connected to the robot body and the rotating platform via a signal line.

[0038] The present invention also proposes a computer-readable storage medium, which stores a computer program, and is characterized in that when the computer program is executed by a processor, it implements the steps of any of the above-mentioned efficient three-dimensional visual positioning methods for large casting grinding.

[0039] The beneficial effects of the present invention are:

[0040] This invention proposes an efficient 3D visual positioning method for polishing large castings. This method addresses the accuracy and efficiency issues of visual positioning for large castings. A 3D industrial camera is mounted on the end of a robot. A rotating platform, controlled by a host computer, rotates to corresponding angles. This method acquires multi-perspective point cloud data of the large casting placed on the platform. A machine vision processing system then performs point cloud registration to reconstruct a 3D point cloud model of the casting, enabling the polishing robot to precisely locate the required spatial position for machining. This method achieves high positioning accuracy and high speed, making it highly valuable for industrial positioning of large castings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The above and other features of the present disclosure will become more apparent through a detailed description of the embodiments shown in conjunction with the accompanying drawings. The same reference numerals in the drawings of the present disclosure represent the same or similar elements. Obviously, the drawings described below are only some embodiments of the present disclosure. It is possible for a person skilled in the art to derive other drawings based on these drawings without inventive effort. In the drawings:

[0042] Figure 1 Shown is a flow chart of the efficient three-dimensional visual positioning method for large casting grinding of the present invention;

[0043] Figure 2 Shown is a schematic structural diagram of the high-efficiency three-dimensional vision platform for large casting grinding of the present invention;

[0044] Figure 3The figure shows the control principle diagram of the high-efficiency three-dimensional visual positioning method for large casting grinding of the present invention;

[0045] Figure 4 The figure shows a flowchart of the global sparse and local dense point cloud registration program of the efficient three-dimensional visual positioning method for large casting grinding of the present invention;

[0046] Figure 5 Shown is a multi-view data diagram of a casting for the efficient three-dimensional visual positioning method for large casting grinding according to the present invention;

[0047] Figure 6 Shown is a casting registration effect diagram of the efficient three-dimensional visual positioning method for large casting grinding of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects of the present invention so as to fully understand the purpose, scheme and effect of the present invention. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other unless there is a conflict. The same reference numerals used throughout the drawings indicate the same or similar parts.

[0049] Reference Figure 1 as well as Figure 3 In Example 1, the present invention proposes an efficient three-dimensional visual positioning method for large casting grinding, comprising the following:

[0050] Step 110: complete the calibration of the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting;

[0051] Step 120: Perform multi-view scanning on the target casting according to a preset single rotation angle and rotation direction, and obtain casting point cloud data of multiple different view angles according to the number of shots when the rotating platform rotates 360 degrees;

[0052] Step 130: Perform multi-view point cloud registration to register the casting point cloud data obtained from multiple different viewpoints into the robot world coordinate system to obtain a globally reconstructed casting point cloud model.

[0053] Step 140: Complete the missing local information on the globally reconstructed casting point cloud model to complete visual positioning.

[0054] In this preferred embodiment, the present invention carries a 3D industrial camera at the end of the robot, which rotates the platform to a corresponding angle under the control of the host computer, and obtains multi-perspective point cloud data of the large casting placed on the platform, and uses the machine vision processing system to perform point cloud registration to complete the reconstruction of the three-dimensional point cloud model of the casting, thereby achieving the spatial position required for the precise positioning of the grinding robot. This method has high positioning accuracy and fast positioning speed. This method is very meaningful for the positioning of large castings in industry. In response to the problems of accuracy and efficiency of visual positioning of large castings, the present invention designs a new scanning platform, and carries a line scanning camera at the end of the robot to capture global geometric information of large castings with fewer perspectives and a large range. On this basis, a secondary high-resolution scan is performed on the local position where the robot needs to perform grinding processing to ensure the positioning efficiency and accuracy of the robot.

[0055] As a preferred embodiment of the present invention, specifically, the calibration of the 3D industrial camera and the rotating platform is completed in the following manner:

[0056] A visual standard plate is placed on the spindle center of the rotating platform, and the 3D industrial camera is focused so as to align the 3D industrial camera with the target casting.

[0057] As a preferred embodiment of the present invention, specifically, the preset single rotation angle is 60 degrees, the rotation direction is forward, and the corresponding number of shots is 6 times.

[0058] Reference Figure 4 As a preferred embodiment of the present invention, specifically, the multi-view point cloud registration process includes the following:

[0059] Calculate the overlap W of any two viewpoint clouds. Assume that the number of 3D points in the two viewpoint clouds is m and n respectively. Use KD tree to detect all similar points in the two viewpoint clouds. Assume that the number of similar points is N. Then the overlap W of the two viewpoint clouds is calculated as follows: It is defined that when W>0.2, the two viewpoint clouds have a certain degree of overlap

[0060] The two viewpoint clouds that meet the overlap requirement W>0.2 are registered. Assume that the two viewpoint clouds that meet the overlap requirement are P={p1,...,p n}, Q={q1,...,q n}, now we want to find a Euclidean transformation R,t such that Constructing the least squares problem When the objective function reaches the minimum, the optimal rotation matrix R and translation matrix t are solved.

[0061] Defining Error Set the maximum number of iterations to 50. When the number of iterations is greater than 50 or e < 0.001, stop the iteration. Otherwise, enter the iteration.

[0062] Output the true value of the rotation and translation matrix and draw the registered image to obtain the globally reconstructed casting point cloud model.

[0063] As a preferred embodiment of the present invention, specifically, the globally reconstructed casting point cloud model is supplemented with local information missing, including the following:

[0064] The local surface of the target casting is perceived twice from different perspectives by an industrial camera to obtain a local dense point cloud model.

[0065] The global model is registered with the local dense model to build a weighted ICP point cloud registration model: Assume that there is P'={p'1,p'2,...,p' n} and Q'={q'1,q'2,...,q' n}Two sets of corresponding points, want to find a Euclidean transformation R,t,

[0066] So that: where ω i >0 is the weight corresponding to each point pair,

[0067] The mathematical model of solving weights is combined with the Welsh function, where the mathematical relationship between the Welsh function and weights is as follows: As for the value of coefficient ν, since there are more registration points and the distance between the points is larger in the initial registration stage, ν = 2 with a smoother curve is selected as the coefficient of the first registration. After each iteration, the proportional coefficient changes to 0.9 times the previous one, that is, ν n =0.9ν n-1 ,

[0068] According to the above, iteratively remove outliers and complete the missing local information.

[0069] Figure 5 The figure shows the multi-view data of the casting obtained by the method of the present invention when it is applied. After being processed by the method of the present invention, it can be obtained Figure 6 The casting registration renderings are shown.

[0070] The present invention also proposes an efficient three-dimensional visual positioning device for large casting grinding, comprising:

[0071] A calibration module is used to calibrate the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting;

[0072] A point cloud data acquisition module is used to perform multi-view scanning on the target casting according to a preset single rotation angle and rotation direction, and obtain casting point cloud data of multiple different perspectives according to the number of shots taken when the rotating platform rotates 360 degrees;

[0073] The point cloud registration module is used to perform multi-view point cloud registration, registering the casting point cloud data obtained from multiple different viewpoints to the robot's world coordinate system to obtain a globally reconstructed casting point cloud model;

[0074] The missing information completion module is used to complete local information missing in the globally reconstructed casting point cloud model and complete visual positioning.

[0075] Reference Figure 2 The present invention also proposes an efficient three-dimensional vision platform for large casting grinding, which applies any of the above-mentioned efficient three-dimensional vision positioning methods for large casting grinding, including:

[0076] The robot body is used to fix the camera at the end and adjust the camera's shooting height and angle, so that the robot can process the casting after positioning;

[0077] A 3D industrial camera is provided at one end of the robot body and is used to capture point cloud data of castings from multiple different perspectives;

[0078] A rotating platform, adapted to the robot body and disposed below the 3D industrial camera, is used to place a target casting and drive the target casting to rotate;

[0079] The host computer is used to complete the calibration work of the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting.

[0080] The target casting is scanned from multiple perspectives according to the preset single rotation angle and rotation direction, and the casting point cloud data of multiple different perspectives is obtained according to the number of shots when the rotating platform rotates 360 degrees.

[0081] Perform multi-view point cloud registration, register the casting point cloud data obtained from multiple different viewpoints to the robot world coordinate system, and obtain a globally reconstructed casting point cloud model.

[0082] The globally reconstructed casting point cloud model is used to complete local information missing and visual positioning;

[0083] The host computer is connected to the robot body and the rotating platform via a signal line.

[0084] When this platform is running,

[0085] 1. Complete the calibration of the 3D industrial camera and the rotating platform. First, use a visual standard plate, place it on the spindle center of the rotating platform, and adjust the focus of the depth camera so that the depth camera is basically aligned with the workpiece.

[0086] 2. Run the multi-view scanning program for the casting. The rotating platform rotates 60° forward in sequence, taking a photo each time it rotates forward until the rotating platform rotates 360°, and then outputs the multi-view point cloud data corresponding to the casting.

[0087] 3. Run the multi-view point cloud registration program. Combined with a highly generalized multi-view point cloud registration algorithm, the obtained point cloud is registered to the robot's world coordinate system to obtain a globally reconstructed casting point cloud model.

[0088] 4. Complete missing local information. Scan the local position point cloud of the casting at high resolution, run the global sparse and local dense point cloud registration program to obtain more accurate casting positioning information.

[0089] The present invention also proposes a computer-readable storage medium, which stores a computer program, and is characterized in that when the computer program is executed by a processor, it implements the steps of any of the above-mentioned efficient three-dimensional visual positioning methods for large casting grinding.

[0090] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

[0091] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing module, or each module may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0092] If the integrated module is implemented in the form of 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, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or system that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0093] Although the present invention has been described in considerable detail and with particularity with respect to several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but rather should be construed as providing a broad possible interpretation of these claims in view of the prior art by reference to the appended claims, thereby effectively encompassing the intended scope of the invention. In addition, the invention has been described above in terms of embodiments foreseen by the inventors for the purpose of providing a useful description, and those insubstantial modifications of the invention that are not currently foreseen may still represent equivalent modifications of the invention.

[0094] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. As long as the technical effects of the present invention are achieved by the same means, they shall fall within the scope of protection of the present invention. Within the scope of protection of the present invention, various modifications and variations of the technical solutions and / or implementation methods may be made.

Claims

1. An efficient 3D visual positioning method for large casting grinding, characterized by: These include: Complete the calibration of the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting; Performing multi-view scanning on the target casting according to a preset single rotation angle and rotation direction, and obtaining casting point cloud data of multiple different view angles according to the number of shots when the rotating platform rotates 360 degrees; Perform multi-view point cloud registration, register the casting point cloud data obtained from multiple different viewpoints to the robot's world coordinate system, and obtain a globally reconstructed casting point cloud model; The globally reconstructed casting point cloud model is used to complete local information missing and visual positioning; Specifically, the globally reconstructed casting point cloud model is supplemented with missing local information, including the following: The local surface of the target casting is perceived twice from different perspectives by an industrial camera to obtain a local dense point cloud model. Register the global model with the local dense model and build a weighted ICP point cloud registration model: Assume that there is and Two sets of corresponding points, want to find a Euclidean transformation , So that: ,in is the weight corresponding to each point pair, The mathematical model of solving weights is combined with the Welsh function, where the mathematical relationship between the Welsh function and weights is as follows: , According to the above method of aligning the global model with the local dense model, outliers are iteratively removed to complete the missing local information.

2. The efficient three-dimensional visual positioning method for large casting grinding according to claim 1 is characterized in that: Specifically, the calibration of the 3D industrial camera and the rotating platform is completed in the following way: A visual standard plate is placed on the spindle center of the rotating platform, and the 3D industrial camera is focused so as to align the 3D industrial camera with the target casting.

3. The efficient three-dimensional visual positioning method for large casting grinding according to claim 1 is characterized in that: Specifically, the preset single rotation angle is 60 degrees, the rotation direction is forward, and the corresponding number of shots is 6 times.

4. The efficient three-dimensional visual positioning method for large casting grinding according to claim 1 is characterized in that: Specifically, the multi-view point cloud registration process includes the following: Calculate the overlap between any two viewpoint clouds , assuming that the number of 3D points in the two viewpoint clouds is and , use KD tree to detect all similar points in the two viewpoint clouds, and assume that their number is , then the overlap of the two viewpoint clouds is The calculation formula is as follows: , The overlap requirement will be met >0.2 to align the two viewpoint clouds, assuming that the two viewpoint clouds that meet the overlap requirement are , Now we want to find a Euclidean transformation , making , construct the least squares problem , when the objective function reaches the minimum, solve the optimal rotation matrix With the translation matrix , Defining Error , set the maximum number of iterations to 50 times. When the number of iterations is greater than 50 or When , stop the iteration, otherwise enter the iteration, Output the true value of the rotation and translation matrix and draw the registered image to obtain the globally reconstructed casting point cloud model.

5. The efficient three-dimensional visual positioning method for large casting grinding according to claim 1 is characterized in that: Specifically, for the coefficient Since the number of registration points in the initial registration stage is large and the distance between the points is large, the curve with a smoother =2 is used as the coefficient for the first registration. After each iteration, the proportional coefficient changes to 0.9 times of the previous one, that is, .

6. An efficient three-dimensional visual positioning device for large casting grinding, characterized by: include: A calibration module is used to calibrate the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting; A point cloud data acquisition module is used to perform multi-view scanning on the target casting according to a preset single rotation angle and rotation direction, and obtain casting point cloud data of multiple different perspectives according to the number of shots taken when the rotating platform rotates 360 degrees; The point cloud registration module is used to perform multi-view point cloud registration, registering the casting point cloud data obtained from multiple different viewpoints to the robot's world coordinate system to obtain a globally reconstructed casting point cloud model; The missing information completion module is used to complete local information missing in the globally reconstructed casting point cloud model and complete visual positioning.

7. An efficient 3D vision platform for large casting grinding, characterized by: The efficient three-dimensional visual positioning method for large casting grinding according to any one of claims 1 to 5 is applied, comprising: The robot body is used to fix the camera at the end and adjust the camera's shooting height and angle, so that the robot can process the casting after positioning; A 3D industrial camera is provided at one end of the robot body and is used to capture point cloud data of castings from multiple different perspectives; A rotating platform, adapted to the robot body and disposed below the 3D industrial camera, is used to place a target casting and drive the target casting to rotate; The host computer is used to complete the calibration work of the 3D industrial camera and the rotating platform so that the 3D industrial camera is aligned with the target casting. The target casting is scanned from multiple perspectives according to the preset single rotation angle and rotation direction, and the casting point cloud data of multiple different perspectives is obtained according to the number of shots when the rotating platform rotates 360 degrees. Perform multi-view point cloud registration, register the casting point cloud data obtained from multiple different viewpoints to the robot world coordinate system, and obtain a globally reconstructed casting point cloud model. The globally reconstructed casting point cloud model is used to complete local information missing and visual positioning; The host computer is connected to the robot body and the rotating platform via a signal line.

8. A computer-readable storage medium storing a computer program, wherein: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images