A method for applying glue and pasting patches to the riser of a sand pot casting based on 3D vision

Abstract

The present invention discloses a method for applying glue and pasting patches to the riser of a sand pot casting based on 3D vision, including the steps of: installing a binocular structured light camera directly above the working field of view vertically; establishing the coordinate system conversion relationship between the robotic arm and the robot; initializing the 3D camera and obtaining the point cloud; preprocessing the point cloud and processing the point cloud; obtaining the pose of the final ring; converting the coordinate system into the robotic arm coordinate system and sending it to the robotic arm for glue application and patch pasting; taking 2D pictures after glue application; sending the counted patch quality to the robotic arm for system judgment; making the patch pasting sequence meet the requirements of processing anti-interference; obtaining the position information of the casting to be patched and glued at the processing position through communication according to the processing progress of the robotic arm. The present invention can realize the automatic glue application at the riser position and the patch pasting application of the riser, replace manual labor, and prevent potential health risks; it can efficiently realize the application of glue application and patch pasting to the sand pot riser under similar working conditions, and at the same time meet the detection of glue application quality, saving space and cost.

Description

Technical Field

[0001] The invention belongs to the technical field of gluing patches on sand pot castings, and in particular relates to a method for gluing patches on the cap of a sand pot casting based on 3D vision. Background Art

[0002] At present, in the application of glue patching in the post-process of factory sand pot castings, the traditional method is to use manual picking, semi-automatic glue coating and patching through manual and mechanical devices, and there are also line laser 3D automatic patching solutions. First of all, the traditional manual patching and placing of workpieces will have the following problems, namely high labor costs, large amounts of dust, obvious odor in the ambient gas, lack of safety guarantees, and low efficiency; while automated production through the 3D mode of line laser will result in the production rhythm not keeping up, low efficiency, and failure to meet production efficiency requirements. The 3D vision proposed in this patent is a method of three-dimensional reconstruction based on the binocular surface structured light method to locate the workpiece. The main hardware involves cameras and projectors. It has high positioning accuracy, can handle the positioning problems of multiple workpieces at the same time, and meet the requirements for gluing quality detection.

[0003] 3D surface structured light camera vision uses computers to achieve the function of preventing human eyes from seeing. It mainly obtains parallax based on image coding information and calculates the depth information of objects using the principle of parallax. Generally, 3D vision systems are composed of computers, image sensors, light sources and digital image processing algorithm programs to obtain the position information of target objects. The identification and positioning of sand pot castings based on 3D vision uses 3D surface structured light cameras for 3D reconstruction, uses point cloud processing technology to identify the blank workpiece to be positioned, and obtains the position information of the blank workpiece in order to cooperate with the robot arm for grasping. This method can automatically complete the gluing and patching of the workpiece, and detect the gluing quality, overcoming the impact of the working environment on the health of workers and the requirement of 24-hour uninterrupted work. Summary of the invention

[0004] The present invention provides a method for gluing and pasting a cap of a sand pot casting based on 3D vision, which can overcome the fact that the sand pot casting gluing and pasting processing process is highly dependent on manpower and overcomes the harsh working environment of manpower; 3D vision technology is used to realize the recognition and positioning of the workpiece, and a mechanical arm is cooperated to complete the gluing and pasting; the whole patch system is in a stable working state and meets the requirements of the gluing quality detection of the patch.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] A method for applying glue to a cap of a sand pot casting based on 3D vision of the present invention comprises the following steps:

[0007] S1. Install the binocular structured light camera vertically above the working field;

[0008] S2. Establish the coordinate system transformation relationship between the robotic arm and the robot;

[0009] S3. Before each gluing, take a 2D picture. Through the TCP / IP robot communication instruction, initialize the camera, trigger the 2D camera to take a picture, then turn off the 2D camera, initialize the 3D camera and obtain the point cloud;

[0010] S4. Point cloud preprocessing: Process the point cloud using processing algorithms such as filtering, segmentation, normal extraction, and plane segmentation to obtain the target point cloud of each patch to be pasted;

[0011] S5. The point cloud extracts the target geometric information through the minimum bounding box, obtains the geometric center of the point cloud by edge extraction and circle center fitting, and performs weighted averaging on several results to suppress the maximum value situation of a single result, obtaining the final pose of the ring;

[0012] S6. Convert the coordinate system to the pose in the robotic arm coordinate system;

[0013] S7. According to the extracted position information of the casting cap, sort the positions so that the patch sequence meets the requirements of processing anti-interference;

[0014] S8. According to the instruction sent by the robot, feedback the patch gluing information of the nth casting, and start the 2D camera to take a picture, recording the imaging picture before gluing;

[0015] S9. After the robotic arm finishes gluing the workpiece, trigger the 2D single camera in the binocular structured light camera to take a picture to obtain the picture, and take the 2D picture after gluing;

[0016] S10. Obtain the image features through the differential processing of the image, analyze the image features, obtain the gluing quality judgment result, and record the gluing quality;

[0017] S11. Paste the patch on the glued cap. According to the recognized number of castings, repeat steps S8 to S11;

[0018] S12. Send the statistically obtained patch quality result to the robotic arm for system judgment;

[0019] Further, the specific steps of S2 include:

[0020] S21. First, establish a robot user-defined coordinate system, and teach the robot user-defined coordinate system with the center of one of the rings as the origin;

[0021] S22. Take a picture to obtain the coordinates of the center of the circle, paste the fiducial point at the end of the robotic arm, and sequentially start taking pictures to obtain the poses of the fiducial points at different positions and record the poses of the robotic arm, establishing the corresponding relationship between the camera coordinate system and the robot user-defined coordinate system;

[0022] S23. At the same time, the deviation of the end of the tool from the flange center is obtained according to the origin marker of the robot user coordinate system for path compensation of the end reaching position of the subsequent robotic arm patching and gluing.

[0023] Further, the specific steps of the S4 step include:

[0024] S41. First, perform a pass-through filter on the point cloud and filter the point cloud according to the Z-value distribution of the working space of the casting;

[0025] S42. Perform point cloud smoothing on the point cloud after the pass-through filter, and use the smoothed point cloud for point cloud segmentation by Euclidean distance;

[0026] S43. Calculate the normal according to the segmented point cloud, and segment the planar ring according to the method of plane segmentation.

[0027] Further, the specific steps of the S5 step include:

[0028] S51. According to the result of the S43 step, the segmented planar ring, establish the minimum bounding box to extract the geometric information of the target - the center pose of the minimum circumscribed box and the centroid of the point cloud;

[0029] S52. Extract the point cloud edge from the point cloud ring segmented by S43, fit the ring according to the edge points, and obtain the center of the point cloud ring;

[0030] S53. Obtain the position of the ring by weighting the center of the minimum bounding box, the centroid of the ring point cloud, and the center extracted from the ring.

[0031] Further, the specific steps of the S10 step include:

[0032] S101. Perform differential processing on the 2D camera images taken in the S9 step and the S8 step to obtain the gluing features;

[0033] S102. Binarize the entire differential image, obtain the gluing features through feature processing of the binary image, analyze the features to determine the quality information of the gluing, and record if it is unqualified.

[0034] The present invention has the following beneficial effects compared with the prior art:

[0035] The present invention applies 3D reconstruction vision technology and point cloud processing technology to the glue coating and patch application of the sand pot cap mouth, enabling automatic glue coating at the cap mouth position, patching of the cap mouth, and completing the application of glue coating quality inspection, replacing manual labor and preventing existing health risks. The present invention uses 3D reconstruction technology to complete the 3D reconstruction of the workpiece at the glue coating and patching station of the cap mouth, obtains a relatively robust target pose through the analysis and processing of the point cloud, and meets the anti-collision requirements of the processing technology by sorting the extracted target information, and can efficiently realize the glue coating and patching application of the sand pot cap mouth under similar working conditions. At the same time, it better meets the detection of glue coating quality, saving space and cost.

[0036] Of course, it is not necessary for any product implementing the present invention to simultaneously achieve all the above-mentioned advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0038] Figure 1 It is a method step diagram of a glue patch for the cap mouth of a sand pot casting based on 3D vision of the present invention;

[0039] Figure 2 It is a specific flowchart of the glue coating quality inspection for step S10 of the present invention;

[0040] Figure 3 It is an installation schematic diagram of the grasping station of the cap mouth of a sand pot casting;

[0041] Figure 4 It is the origin of the user-defined coordinate system with the cross center of the circular marker as the center for the robotic arm in step S21 of the present invention;

[0042] Figure 5 It is one of the real scene pictures of the glue patch for the cap mouth of a sand pot casting;

[0043] Figure 6 It is the second real scene picture of the glue patch for the cap mouth of a sand pot casting. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0045] Please refer to Figures 1-6 As shown in the figure, a method for applying glue and pasting patches to the riser of a sand pot casting based on 3D vision according to the present invention includes the following steps:

[0046] S1. Install the binocular structured light camera directly above the working vision vertically;

[0047] S2. Establish the coordinate system transformation relationship between the robotic arm and the robot, which is specifically achieved through the following methods;

[0048] S21. First, establish the robot user-defined coordinate system, and teach the robot user-defined coordinate system with the center of one of the rings as the origin; in this specific embodiment, the cross center of the ring marker is used as the origin, and the 3D camera is started to take pictures;

[0049] S22. Take pictures to obtain the coordinates of the center of the circle, paste the marker points at the end of the robotic arm, and start taking pictures in sequence to obtain the poses of the marker points at different positions and record the poses of the robotic arm, and establish the corresponding relationship between the camera coordinate system and the robot user-defined coordinate system; specifically, it is achieved through the following methods: paste the ring marker at the end of the robotic arm, take pictures within the camera's field of view, move N points (N>9) and record the poses of the robotic arm, and at the same time take pictures with the 3D camera each time it moves to obtain the poses in the camera coordinate system; use the least squares method to calculate the rigid transformation matrix of the two coordinate systems and establish the corresponding relationship, where R is the rotation matrix and T is the translation parameter;

[0050]

[0051] S23. At the same time, calculate the deviation of the end of the tool from the center of the flange according to the origin marker of the robot user coordinate system for subsequent path compensation of the robotic arm for pasting and applying glue; specifically, use the pose of the 3D camera coordinate system of the origin marker point to calculate the deviation of the end of the marker point during the movement, and compensate the deviation in the subsequent calculation to obtain the actual movement pose of the robotic arm.

[0052] S3. Take a 2D picture before each application of glue. Through the TCP / IP robot communication instruction, initialize the camera, trigger the 2D camera to take pictures, then turn off the 2D camera, initialize the 3D camera and obtain the point cloud; specifically, perform point cloud filtering processing to obtain a smooth target area, perform Euclidean distance segmentation, and obtain the point cloud of the sand pot according to the number and size of the point cloud clusters;

[0053] S4. Point cloud preprocessing: Process the point cloud by means of filtering, segmentation, normal extraction, plane segmentation, etc., which is specifically achieved through the following steps:

[0054] S41. First, perform a pass-through filter on the point cloud to filter the point cloud according to the Z-value distribution of the working space of the casting;

[0055] S42. Smooth the point cloud with a thick direct filter, and perform point cloud segmentation based on the Euclidean distance using the smoothed point cloud;

[0056] S43. Calculate the normal vectors based on the segmented point cloud, and segment the planar ring according to the plane segmentation method.

[0057] S5. Extract the target geometric information of the point cloud through the minimum bounding box, and obtain the center of the point cloud by edge extraction and circle center fitting; it is specifically implemented through the following steps;

[0058] S51. Based on the result of step S43, the segmented planar ring, establish a minimum bounding box to extract the geometric information of the target - the center pose of the minimum circumscribed box and the centroid of the point cloud:

[0059] S52. Extract the point cloud edge from the point cloud ring segmented by S43, fit the circle according to the edge points, and obtain the center of the point cloud ring;

[0060] S53. Obtain the position of the ring by weighting the center of the minimum bounding box, the centroid of the ring point cloud, and the center extracted from the ring;

[0061] Among them, specifically, according to the point cloud segmented by Euclidean clustering, extract the plane of the sand can cap, extract the minimum circumscribed bounding box through the ring point cloud, obtain the geometric center and centroid, and at the same time extract the point cloud edge, and obtain the ring center through fitting;

[0062] P = (X, Y, Z)

[0063] P = a * P1 + b * P2 + c * P3 (a < 1; b < 1; c < 1; a + b + c = 1)

[0064] To suppress the maximum value situation of a single result, several results are weighted and averaged to obtain the final pose of the ring;

[0065] S6. Convert the coordinate system into the robotic arm coordinate system and send it to the robotic arm for gluing and pasting, and then turn off the camera;

[0066] S7. At the same time, trigger the 2D photographing in the 3D camera to obtain pictures, and take 2D pictures after gluing; wait for the robotic arm to send an instruction to obtain the workpiece index to be pasted and glued, send the position data, and at the same time start the 2D camera to take pictures, then the camera glues, waits for the gluing completion instruction, and takes pictures with the 2D camera after completion;

[0067] S8. Record the gluing quality by obtaining the image features through the differential processing of the images; perform differential processing on the photos before and after gluing to obtain the gluing features, analyze the gluing features, and evaluate whether the gluing quality is qualified. It is specifically implemented through the following steps:

[0068] S81. Perform differential processing on the 2D camera images taken in step S7 and step S3 to obtain the gluing features.

[0069] S82. Binarize the entire differential image, obtain the gluing features through feature processing of the binary image, analyze the features to determine the quality information of the gluing. If it is a non-compliant record, proceed to the next step.

[0070] S9. Patch the glued cap openings. According to the recognized number of castings, repeat steps S3 to S9, and send the statistically obtained patching quality to the robotic arm for system judgment. Specifically, determine the number of cap opening patches in the same batch according to the number of point cloud Euclidean clustering segmentations. After completion, send the number of unqualified gluing to the robotic arm for system judgment.

[0071] S10. Obtain image features through differential processing of the image, analyze the image features to obtain the gluing quality judgment result, and record the gluing quality. According to the extracted position information of the cast cap openings, sort the positions so that the patching order meets the requirements of processing anti-interference; that is, by sorting the target point set to meet the requirements of the gluing and patching processes.

[0072] S11. Patch the glued cap openings. According to the recognized number of castings, repeat steps S8 to S11.

[0073] S12. Send the statistically obtained patching quality result to the robotic arm for system judgment, and feedback the patching and gluing information of the nth casting according to the instructions sent by the robot.

[0074] The present invention applies 3D reconstruction vision technology and point cloud processing technology to the gluing and patching of sand pot cap openings, can realize automatic gluing of cap opening positions and patching applications of cap openings, replace manual labor, and prevent existing health risks; the present invention uses 3D reconstruction technology to complete the 3D reconstruction of workpieces at the cap opening gluing and patching stations, and obtains a relatively robust target pose through the analysis and processing of point clouds. By sorting the extracted target information, it meets the anti-collision requirements of the processing technology, can efficiently realize the application of gluing and patching of sand pot cap openings under similar working conditions, and at the same time better meets the detection of gluing quality, saving space and cost.

[0075] The above-disclosed preferred embodiments of the present invention are only used to help explain the present invention. The preferred embodiments do not describe all details in detail, nor limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principle and practical application of the present invention, so that those skilled in the relevant technical field can understand and utilize the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A method for applying glue and pasting patches to the riser of a sand pot casting based on 3D vision, characterized in that, It includes the following steps: S1. Install the binocular structured light camera directly above the working field of view vertically; S2. Establish the coordinate system transformation relationship between the robotic arm and the robot; S3. Before each gluing, take a 2D picture, initialize the camera through the TCP / IP robot communication instruction, trigger the 2D camera to take a picture, then turn off the 2D camera, initialize the 3D camera and obtain the point cloud; S4. Point cloud preprocessing: Process the point cloud with filtering, segmentation, normal extraction, and plane segmentation processing algorithms to obtain the target point cloud of each patch to be pasted; S5. Extract the target geometric information from the point cloud through the minimum bounding box, obtain the geometric center of the point cloud by edge extraction and fitting the center of the circle, and perform weighted averaging of several results to suppress the maximum value situation of a single result, obtaining the pose of the final circular ring; The specific steps of S5 include: S51. According to the result of step S4, segment the planar circular ring, and establish the minimum bounding box to extract the geometric information of the target - the center pose of the minimum circumscribed box and the centroid of the point cloud; S52. Extract the edge of the point cloud from the segmented point cloud circular ring by S4, fit the circular ring according to the edge points, and obtain the center of the point cloud circular ring; S53. Obtain the position of the circular ring by weighting the center of the minimum bounding box, the centroid of the circular - ring point cloud, and the center extracted from the circular ring; S6. Convert the coordinate system into the pose in the robotic - arm coordinate system; S7. According to the extracted position information of the casting cap, sort the positions so that the patch - pasting order meets the requirements of processing anti - interference; S8. According to the instruction sent by the robot, feedback the patch - pasting and gluing information of the nth casting, and start the 2D camera to take a picture to record the imaging picture before gluing; S9. After the robotic arm finishes gluing the workpiece, trigger the 2D single camera in the binocular structured light camera to take a picture to obtain the picture, and take the 2D picture after gluing; S10. Obtain the image features through the differential processing of the image, analyze the image features, obtain the gluing quality judgment result, and record the gluing quality; S11. Paste patches on the glued cap, and according to the recognized number of castings, repeat steps S8 to S11; S12. Send the statistically obtained patch - pasting quality result to the robotic arm for system judgment.

2. A method for gluing and pasting the riser of a sand pot casting based on 3D vision according to claim 1, characterized in that, The specific steps of S2 include: S21. First, establish the robot user - defined coordinate system, and teach the robot user - defined coordinate system with the center of a circular ring as the origin; S22. Take a picture to obtain the coordinates of the center of the circle, paste the fiducial points at the end of the robotic arm, start taking pictures in sequence to obtain the poses of the fiducial points at different positions and record the poses of the robotic arm, and establish the corresponding relationship between the camera coordinate system and the robot user - defined coordinate system; S23. At the same time, obtain the deviation of the distance from the tool end to the flange center according to the origin marker of the robot user coordinate system for path compensation of the end - reaching position of subsequent robotic - arm patch - pasting and gluing.

3. A method for applying glue and pasting patches to the riser of a sand pot casting based on 3D vision according to claim 1, characterized in that, The specific steps of S4 include: S41. First, perform a pass - through filter on the point cloud, and filter the point cloud according to the Z - value distribution of the working space of the casting; S42. Perform point - cloud smoothing processing on the processed point cloud, and use the smoothed point cloud for point - cloud segmentation by Euclidean distance; S43. Calculate the normal according to the segmented point cloud, and segment the planar circular ring according to the method of plane segmentation.

4. A method for applying glue and pasting patches to the riser of a sand pot casting based on 3D vision according to claim 1, characterized in that, The specific steps of S10 are as follows: S101. Perform differential processing on the 2D camera images taken in steps S9 and S8 to obtain the gluing features; S102. Binarize the entire differential image, obtain the gluing features through feature processing of the binary image, analyze the features to determine the quality information of the gluing, and record if it is irregular.

Images