Stacked frame unstacking grabbing method based on 3D vision, grabbing system and grabbing device

Abstract

The application discloses a 3D vision-based stacking frame unstacking grabbing method, a grabbing system and a grabbing device, which comprise a single-layer frame grabbing method in a stacking frame, a stacking frame unstacking grabbing method, a stacking frame unstacking grabbing system configured to realize the unstacking grabbing method, and a stacking frame unstacking grabbing device configured to realize the unstacking grabbing method. The application can accurately complete the identification and grabbing of different frames and different isolation papers, solves the problem of material falling during grabbing, and can simultaneously grab multiple frames for unstacking operation, thereby improving the frame feeding efficiency.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of photovoltaic module production, in particular to a stacking frame unstacking grabbing method, a grabbing system and a grabbing device based on 3D vision. BACKGROUND

[0002] In the production process of photovoltaic modules, an edge frame needs to be assembled around the glass module. The four edge frames are divided into two long edge frames and two end edge frames. The long edge frame and the short edge frame are both stacked in multiple layers, and each layer is stacked with multiple edge frames. Since the front and back surfaces of the edge frame can be concave-convex matched, the edge frame is stacked in layers with the front surface of one layer facing up and the back surface of the other layer facing up. When two layers of edge frames are stacked, the four sides are flat. Therefore, the edge frame can be stacked in multiple layers with multiple edge frames in each layer. In order to prevent friction between the layers of edge frames, edge frame isolation paper is provided between each two layers of edge frames. In order to prevent the two ends of the edge frame from being scratched, corner code isolation paper is provided at the two ends of the edge frame. Therefore, during the feeding process of the edge frame, the edge frame isolation paper and the corner code isolation paper need to be grabbed first, and then the multiple front surface edge frames and back surface edge frames are grabbed and transported to the processing station. After the edge frames in the same layer are taken out multiple times, the next layer of edge frames is fed. The edge frame isolation paper and the corner code isolation paper need to be taken out first. Therefore, the grabbing system needs to be able to identify the edge frame isolation paper, the corner code isolation paper, the front surface edge frame and the back surface edge frame. In order to improve the grabbing efficiency, the grabbing device needs to grab multiple edge frames at the same time.

[0003] In the prior art, for example, a kind of unstacking method, system, device, equipment and storage medium disclosed in Chinese patent publication No. CN118343510A. In this scheme, it is judged whether the top layer of the goods stack is full stack layer;If the top layer is full stack layer, the top layer pose is determined according to the package point cloud of each package, and the template pose matched with the top layer pose is selected from the preset template pose as the first pose;Then, based on the identified package point cloud, the grabbing pose for grabbing all the packages included in the top layer is determined;Since each template pose corresponds to a preset grabbing mode, the mechanical arm can be controlled to grab all the packages included in the top layer from the goods stack to the specified position according to the grabbing pose and the grabbing mode corresponding to the first pose, so that the mechanical arm can grab one layer of packages at a time during unstacking;Although the shooting of 3D camera, the identification of point cloud and the judgment of package pose are also applied in this scheme, the above scheme takes out one layer of packages at a time. Since the spacing between adjacent edge frames during stacking of photovoltaic edge frames does not meet the requirements of the grabbing device, the edge frame is taken out at a time, which may cause the problem of falling edge frame during grabbing.

[0004] Therefore, it is necessary to provide a 3D vision-based stack frame unstacking grabbing method, a grabbing system and a grabbing device to solve the above technical problems. SUMMARY

[0005] The main purpose of the present application is to provide a 3D vision-based stack frame single-layer frame grabbing method, which can accurately identify and grab different frames, solve the problem of frame dropping during grabbing, and simultaneously grab multiple frames for unstacking operation, thereby improving the efficiency of frame feeding.

[0006] The present application achieves the above-mentioned purpose by the following technical scheme: a 3D vision-based stack frame single-layer frame grabbing method, comprising the following steps:

[0007] S1, the maximum number of frames grabbed by the frame grabbing fixture at one time is m, the number of frames in a single-layer frame is M, and all the frames in the single-layer frame are divided into n frame units according to m, n = [M / m] + 1;

[0008] S2, a 3D camera is used to take a photo of the set center area of the first frame unit, collect point cloud data of m frames, and obtain frame point cloud;

[0009] S3, the frame point cloud is preprocessed, then matched with the set frame point cloud model, the category information and the pose information of the frame are obtained, and the spacing between every two adjacent frames in the first frame unit 1~m is calculated; the category information of the frame includes front frame and back frame;

[0010] S4, whether the spacing between every two adjacent frames in 1~m frames meets the set spacing range is judged, if all meet, 1~m frames are grabbed according to the category information and the pose information obtained in step S3, the grabbing of the first frame unit is completed, and step S7 is executed; otherwise, steps S5~S6 are executed;

[0011] S5, the frame combination that does not meet the set spacing range is determined, and is recorded as frame combination i = {frame A i , frame B i}, that is, the spacing between adjacent frame A i and frame B i does not meet the set spacing range, i = 1, 2, …, p, 1≤p<m-1, p is the number of frame combinations determined not to meet the set spacing range;

[0012] S6, according to the category information and the pose information obtained in step S3, the first 1~frame A1 is grabbed for the first time, then frame B1~frame A2 is grabbed for the second time, and so on, until frame B p ~mth frame is grabbed, and the grabbing of the first frame unit is completed;

[0013] S7, after completing the grabbing of all the frame units in steps S2-S6, the grabbing of the single-layer frame is completed.

[0014] Another main purpose of the present application is to provide a 3D vision-based de-stacking and grabbing method for stacked frames, which can accurately identify and grab different frames and different isolation papers, and solve the problem of material falling during grabbing.

[0015] The present application achieves the above-mentioned purposes through the following technical solutions: a 3D vision-based de-stacking and grabbing method for stacked frames, comprising the following steps:

[0016] S100, the 3D camera takes a picture of the top layer of the whole stack of frames to determine the overall height of the whole stack of frames;

[0017] S200, the 3D camera takes a picture of the four corners of the whole stack of frames to determine the center of the whole stack of frames and correct the stacking position of the whole stack of frames;

[0018] S300, grabbing of the frame isolation paper:

[0019] S301, the 3D camera moves above the frame isolation paper to take a picture of the frame isolation paper, collects the point cloud of the frame isolation paper, and obtains the frame isolation paper point cloud;

[0020] S302, the collected frame isolation paper point cloud is preprocessed, the preprocessed frame isolation paper point cloud is matched with a preset frame isolation paper point cloud model, and the pose information of the frame isolation paper point cloud is obtained;

[0021] S303, according to the pose information of the frame isolation paper obtained in step S302, a corresponding grabbing clamp is selected, the frame isolation paper is grabbed and placed into the isolation paper box;

[0022] S400, grabbing of the single-layer frame: according to the single-layer frame grabbing method as described in claim 1, the grabbing of the single-layer frame is completed;

[0023] S500, during the grabbing of the upper layer frame, the height of the layer frame is further obtained when taking a picture of the frame unit, the 3D camera is adjusted to the best picture height according to the height of the upper layer frame, and the correction of the 3D camera picture height is completed;

[0024] S600, repeat steps S300-S500 until the whole stack of frames is taken, and the de-stacking of the stacked frames is completed.

[0025] Another purpose of the present application is to provide a 3D vision-based de-stacking and grabbing system for stacked frames, which is used to complete the above-mentioned de-stacking and grabbing method for stacked frames, and comprises

[0026] The centering photographing correction module photographs the four corners of the whole stack of frames, determines the center of the whole stack of frames, and performs stack position correction on the whole stack of frames.

[0027] The point cloud acquisition module is configured to acquire point cloud data and depth information of the frame and the separator paper.

[0028] The point cloud data processing module is configured to pre-process the point cloud data and the depth information acquired by the point cloud acquisition module.

[0029] The point cloud matching module is configured to match the pre-processed separator paper point cloud with a pre-set separator paper point cloud model to acquire category information and pose information of the separator paper, the category information of the separator paper being frame separator paper and corner code separator paper; match the pre-processed frame point cloud with a pre-set frame point cloud model to acquire type information and pose information of the frame, and calculate the distance between each adjacent two frames in the plurality of frames and determine whether the distance meets the requirements.

[0030] The pose conversion module is configured to convert the pose information of the frame point cloud and the separator paper point cloud into grasping pose information of a robot gripper by using a pose conversion matrix obtained by hand-eye calibration.

[0031] The grasping device module is configured to receive control instructions and pose data output by the pose conversion module and grasp the separator paper and the frames in sequence.

[0032] The photographing height correction module is configured to further acquire the height of the frame in the frame unit during photographing of the frame in the upper frame grasping process, adjust the 3D camera to the optimal photographing height according to the height of the upper frame, and complete the correction of the photographing height of the 3D camera.

[0033] Another object of the present application is to provide a 3D vision-based de-stacking and grasping device for frames, which is used to complete the above-mentioned de-stacking and grasping method for frames, and comprises a mounting plate arranged at the active end of a mechanical hand, a frame separator paper grasping clamp, a corner code separator paper grasping clamp and a frame grasping clamp arranged on the mounting plate respectively.

[0034] Further, the frame separator paper grasping clamp comprises a first air cylinder, a moving frame driven by the first air cylinder to move up and down, and a plurality of first suction cups fixed to the lower end of the moving frame and used to suck the frame separator paper.

[0035] Further, the corner code separator paper grasping clamp comprises a second air cylinder and a pair of clamping jaws driven by the second air cylinder to perform opening or clamping actions and used to clamp the corner code separator paper.

[0036] Further, the frame clamping clamp comprises a plurality of air path blocks installed on the mounting plate at equal intervals along a first direction, a plurality of second suction cups fixedly arranged along a second direction at lower ends of the air path blocks, and a motor for driving the plurality of air path blocks to simultaneously increase or decrease the intervals along the first direction, the first direction being perpendicular to the second direction, the first direction being the width direction of the frame and the second direction being the length direction of the frame when the frame clamping clamp sucks the frame.

[0037] Compared with the prior art, the beneficial effects of the stack frame unstacking and grabbing method, the grabbing system and the grabbing device based on 3D vision of the application are as follows:

[0038] (1) Before grabbing, the categories and poses of the frame and the isolation paper can be identified respectively according to the point cloud data collected by the 3D camera, the spacing information between each two adjacent frames in the multiple frames is calculated, and it is judged whether the spacing between the two adjacent frames meets the requirements; the frames with the spacing meeting the requirements are taken away in turn, and the grabbing clamp can take away multiple frames each time, which can improve the efficiency of frame unstacking;

[0039] (2) The grabbing system is provided with a photographing height correction module and a centering photographing correction module, the photographing height correction module can output the frame stack position height information according to the last collected frame point cloud, adjust the 3D camera to the best photographing height, correct the photographing height of the whole stack of frames, and ensure that the photographing field height of the 3D camera is within the range; the centering photographing correction module, the 3D camera photographs the four corners of the whole stack of frames, determines the center of the whole stack of frames, corrects the stack position of the whole stack of frames, can ensure that the frame can enter the photographing field, and can photograph the center area of the frame, and ensures the accuracy of the photographing positioning;

[0040] (3) The frame isolation paper grabbing clamp, the corner code isolation paper grabbing clamp and the frame grabbing clamp are integrated on one robot, which can save the robot, and can quickly select different clamping clamps according to the grabbing needs, and can improve the grabbing pace.

[0041] (4) When matching the collected point cloud with the pre-constructed point cloud model, first, the point cloud is roughly matched based on the fast point feature histogram (FPFH), and then the pose result is fine-tuned by using the iterative closest point algorithm (ICP), the rough matching can provide a reliable initial position for the precise matching of the iterative closest point algorithm (ICP), thereby speeding up the registration speed of the iterative closest point algorithm (ICP), realizing the precise positioning of the frame and the isolation paper, ensuring the accuracy of the grabbing position, and solving the problem of frame dropping due to low grabbing positioning accuracy.

DRAWINGS

[0042] Figure 1A three-dimensional structural schematic diagram of the grabbing device of the embodiment of the present application;

[0043] Figure 2 A three-dimensional structural schematic diagram of the grabbing device of the embodiment of the present application;

[0044] Figure 3 A three-dimensional structural schematic diagram of the frame grabbing clamp of the embodiment of the present application;

[0045] Figure 4 A three-dimensional structural schematic diagram of the isolation paper grabbing clamp of the embodiment of the present application;

[0046] Figure 5 A step of the unstacking grabbing method of the stacked frame of the embodiment of the present application;

[0047] The numbers in the figure represent:

[0048] 3-grabbing device, 31-mounting plate, 32-frame grabbing clamp, 321-air path block, 322-second suction cup, 323-motor, 33-frame isolation paper grabbing clamp, 331-first air cylinder, 332-moving frame, 333-first suction cup, 34-corner code isolation paper grabbing clamp, 341-second air cylinder, 342-clamping jaw, 35-3D camera, 36-laser range finder.

DETAILED DESCRIPTION

[0049] In the process of frame unstacking, the frame isolation paper and the corner code isolation paper need to be grabbed first, and then the multiple front frames and back frames are grabbed and transported to the processing station. After the frames of the same layer are taken out multiple times, the frame isolation paper and the corner code isolation paper need to be taken out before the next layer of frame is fed. Therefore, in the process of frame feeding, the isolation paper on the frame needs to be taken out before the frame can be transported. Therefore, in the process of frame feeding, the isolation paper and the frame need to be grabbed. Therefore, our grabbing device can grab the frame isolation paper, the corner code isolation paper and the frame at the same time. Correspondingly, as shown in the figure, the unstacking grabbing device 3 of the stacked frame based on 3D vision includes a manipulator, a mounting plate 31 arranged at the active end of the manipulator, a frame isolation paper grabbing clamp 33, a corner code isolation paper grabbing clamp 34 and a frame grabbing clamp 32 arranged on the mounting plate 31 respectively. Figures 1-4

[0050] The frame isolation paper grabbing clamp 33 includes a first air cylinder 331, a moving frame 332 driven by the first air cylinder 331 to move up and down, and a plurality of first suction cups 333 fixed at the lower end of the moving frame 332 and used to suck the frame isolation paper.

[0051] The corner code isolation paper grabbing clamp 34 includes a second air cylinder 341 and a pair of clamping jaws 342 driven by the second air cylinder 341 to perform opening or clamping actions and used to clamp the corner code isolation paper.​

[0052] The frame clamping fixture 32 comprises a plurality of air path blocks 321 installed on the mounting plate 31 at equal intervals along a first direction, a plurality of second suction cups 322 fixedly arranged at the lower end of each air path block 321 along a second direction, and a motor 323 driving the plurality of air path blocks 321 to simultaneously increase or decrease the interval along the first direction, the first direction being perpendicular to the second direction, and the first direction being the width direction of the frame and the second direction being the length direction of the frame when the frame is sucked. In order to ensure the stability of the grabbing, the second suction cup 322 is adsorbed on the center line of the short side of the frame. If the adsorbed position deviates from the center line of the short side, the frame may fall off.

[0053] The motor 323 drives the plurality of air path blocks 321 to simultaneously move along the first direction, which can simultaneously decrease or increase the interval between the plurality of air path blocks 321, so as to adjust the interval of the plurality of second suction cups 322 along the first direction, so as to be suitable for the synchronous adsorption of a plurality of frames with different widths. This not only improves the efficiency of carrying, but also has higher universality and better flexibility. In the embodiment, ten air path blocks 321 are arranged at equal intervals along the first direction, which can simultaneously adsorb ten frames, thereby improving the adsorption efficiency. In other embodiments, the number of air path blocks 321 can be adjusted according to actual conditions, which is not limited here. The plurality of second suction cups 322 fixedly arranged at the lower end of each air path block 321 along the second direction are simultaneously adsorbed on the same frame, which can improve the accuracy and reliability of adsorption.

[0054] In order to ensure the accuracy and reliability of the frame adsorption, the frame clamping fixture 32 is provided with two groups and is arranged at the two ends of the mounting plate 31, and the two groups of frame clamping fixtures 32 are adsorbed at the two ends of the frame. The mounting plate 31 is provided with a 3D camera 35 and a laser range finder 36, and the 3D camera 35 and the laser range finder 36 are arranged between the two groups of frame clamping fixtures 32. In the embodiment, the 3D camera 35 is a laser galvanometer camera, and the laser range finder 36 measures the height and the distance between planes.

[0055] When the frame is unstacked, the unstacking of each layer is completed first, and then the unstacking of the entire stack of frames is completed. Here, we first explain the single-layer frame unstacking method. The single-layer frame refers to a layer of front frame and a layer of back frame. The front frame is taken out first, and then the back frame is taken out. The grabbing method of the front frame and the back frame is the same. Based on the 3D vision, the single-layer frame grabbing method of the stacking frame is as follows:

[0056] S1, the maximum number of edge frames grabbed by the edge frame grabbing clamp at one time is m, the number of edge frames in a single layer of edge frames is M, and all the edge frames in the single layer of edge frames are divided into n edge frame units according to m, n = [M / n] + 1; wherein n is an upward integer; the M edge frames are arranged in sequence;

[0057] S2, the set center area of the first edge frame unit is photographed by using a 3D camera, point cloud data of the m edge frames is collected, and edge frame point cloud is obtained;

[0058] The edge frame grabbing clamp can grab at most m edge frames. Since the spacing between the edge frames may increase when the edge frames are arranged, in order to ensure that as many edge frames as possible are grabbed, the 3D camera needs to increase the photographing field of view when photographing. The photographing field of view needs to be greater than the field of view range of the m edge frames. Therefore, in order to facilitate calculation and understanding, in the embodiment, the photographing field of view of the 3D camera is calculated by the field of view range occupied by k edge frames, wherein m < k, and at the same time, it can be ensured that when all the edge frames are offset to the maximum, at most m edge frames can still be grabbed. In other embodiments, other ways can be used for calculation, as long as the photographing field of view range of the 3D camera can ensure that when all the edge frames are offset to the maximum, the point cloud of m edge frames can still be collected.

[0059] S3, the edge frame point cloud is preprocessed, and then matched with a set edge frame point cloud model to obtain category information of the edge frames and pose information of the edge frames, and the spacing between every two adjacent edge frames in the first 1-m edge frames of the first edge frame unit is calculated; the category information of the edge frames includes front edge frames and back edge frames;

[0060] Specifically, in the S3 step, the categories of the edge frames include front edge frames and back edge frames. Since the front and back surfaces of the edge frames can be matched concave-convex, the edge frames are stacked in turn according to the edge frames with the front surface upward and the edge frames with the back surface upward. When the two layers of edge frames are stacked, the four side edges are exactly planar, so the edge frames can be stacked in multiple layers and multiple roots in each layer to form a stack. The edge frames include long edge frames and short edge frames. Two long edge frames and a short edge frame are needed to assemble each glass assembly. The long edge frames and the short edge frames are distinguished as front edge frames and back edge frames, and the stacking method of the long edge frames and the short edge frames is the same and the grabbing method is the same.

[0061] When the collected bounding box point cloud is matched with the bounding box point cloud model, the collected bounding box point cloud needs to be preprocessed first, the preprocessing including denoising, filtering, down-sampling, etc., to improve the efficiency and accuracy of subsequent processing; after the preprocessing is completed, the point cloud is coarsely matched based on the fast point feature histogram (FPFH) to obtain initial pose information, and then the initial pose information is fine-tuned by using the iterative closest point algorithm (ICP), and the category information of the bounding box and the pose information of the bounding box point are outputted after the fine-tuning. The coarse matching can provide a reliable initial position for the precise matching of the iterative closest point algorithm (ICP), thereby accelerating the registration speed of the iterative closest point algorithm (ICP) and realizing the rapid positioning of the bounding box. After the iterative closest point algorithm (ICP) matching, the specific position of the bounding box in the point cloud data can be obtained, including the center point coordinates of the object, the attitude (such as the rotation angle), etc.

[0062] Before the bounding box is grabbed, the bounding box point cloud model needs to be constructed, and the construction method of the bounding box point cloud model is as follows:

[0063] S31, scanning a single bounding box to obtain a bounding box point cloud set;

[0064] S32, preprocessing the bounding box point cloud set, the preprocessing including denoising, filtering, down-sampling, etc.;

[0065] S33, identifying and saving the type of the bounding box, and labeling the grabbing point in the bounding box point cloud set to obtain the bounding box point cloud model, and saving the bounding box point cloud model.

[0066] Since the air path blocks 321 on the grabbing clamp are arranged at equal intervals in the first direction when the grabbing clamp is grabbing, the second suction cups 322 are also arranged at equal intervals in the first direction, so that when the bounding boxes are sucked, the interval between the center lines of the short sides of the adjacent two bounding boxes to be sucked is consistent with the interval of the second suction cups 322 in the first direction. If the interval between a certain bounding box and the adjacent bounding box is too small or too large and does not meet the interval requirement, the position is abnormal, which will cause a number of second suction cups 322 to be unable to be adsorbed on the center line of the short side of the bounding box, resulting in a deviation of the adsorption position and a situation of dropping the bounding box, affecting the bounding box grabbing operation. Therefore, the interval information between every two adjacent bounding boxes in the bounding box unit needs to be obtained to confirm whether the interval between every two adjacent bounding boxes meets the requirement, and the grabbing clamp adopts different grabbing modes according to the interval information.

[0067] S4, judging whether the interval between every two adjacent bounding boxes in the 1st to mth bounding box meets the set interval range, if all meet the set interval range, grabbing the 1st to mth bounding box according to the category information and the pose information obtained in step S3 to complete the grabbing of the first bounding box unit, and executing step S7; otherwise, executing steps S5-S6;

[0068] S5, determine the frame combination that does not meet the set spacing range, and mark it as frame combination i={frame A i , frame B i}, that is, the spacing between adjacent frame A i and frame B i does not meet the set spacing range, i=1, 2, …, p, 1≤p

[0069] S6, according to the category information and pose information obtained in step S3, first grasp the 1st~frame A1, and then secondly grasp the frame B1~frame A2, and so on, until the frame B p ~mth frame is grasped, completing the grasping of the first frame unit;

[0070] S7, after completing steps S2-S6, grasp all subsequent frame units to complete the single-layer frame grasping.

[0071] Because there is a separation paper between the upper and lower single-layer frames, when grasping the whole stack of frames, the frame separation paper needs to be removed before the single-layer frame is grasped. Therefore, the photovoltaic frame unstacking method based on 3D vision, as shown in Figure 5 , has the following steps:

[0072] S100, the 3D camera takes a photo of the top layer of the whole stack of frames to determine the overall height of the whole stack of frames. Because the overall height of the whole stack of frames is inconsistent due to different layers of incoming whole stack of frames, the overall height of the whole stack of frames needs to be confirmed by taking a photo with the 3D camera first, so as to determine the optimal shooting height, ensure that the shooting field height of the 3D camera is within the range, and make the shooting clear and facilitate the collection of subsequent point clouds.

[0073] S200, the 3D camera takes a photo of the four corners of the whole stack of frames to determine the center of the whole stack of frames and correct the stack position of the whole stack of frames. Because the whole stack of frames may be offset when placed in the feeding area, if the 3D camera directly takes a photo to collect frame point clouds, it will cause the photo to be offset, which may cause some frames not to enter the shooting field or not to be able to take a photo of the center area of the frame. By taking a photo of the four corners of the whole stack of frames with the 3D camera, the center of the whole stack of frames is determined. By determining the center position of the whole stack of frames, the 3D camera can be moved above the corresponding frame to take a photo, which can ensure that the frame enters the shooting field and can take a photo of the center area of the frame, ensuring accurate photo positioning.

[0074] S300, grasping of the frame separation paper, including the following steps:

[0075] S301, the 3D camera moves above the frame separation paper to take a photo of the separation paper, collects the point cloud of the frame separation paper, and obtains the frame separation paper point cloud.

[0076] S302, pre-process the collected frame isolation paper point cloud, match the pre-processed frame isolation paper point cloud with a preset frame isolation paper point cloud model, and obtain the pose information of the frame isolation paper point cloud;

[0077] Specifically, in the S302 step, the pre-processing of the frame isolation paper point cloud includes denoising, filtering, down-sampling, etc. Before the frame isolation paper is grabbed, a frame isolation paper point cloud model needs to be constructed. The construction method of the frame isolation paper point cloud model is as follows:

[0078] S3021, scan the frame isolation paper to obtain a frame isolation paper point cloud set;

[0079] S3022, pre-process the frame isolation paper point cloud set, which includes denoising, filtering, down-sampling, etc.

[0080] S3023, identify and save the type of the frame isolation paper, mark the grabbing point in the frame isolation paper point cloud set, obtain the frame isolation paper point cloud model, and save the frame isolation paper point cloud model.

[0081] S303, according to the output pose information of the frame isolation paper, select the corresponding grabbing fixture, grab the frame isolation paper and put it into the isolation paper box;

[0082] S400, single-layer frame grabbing, according to the single-layer frame grabbing method in the above steps S1-S7, the single-layer frame grabbing is completed. Here, the single-layer frame refers to one layer of front frame and one layer of back frame. The front frame is taken out first, and then the back frame. The grabbing method of the front frame and the back frame is the same, so it is not repeated.

[0083] S500, in the upper frame grabbing process, further obtain the height of the layer of frame when taking a photo of the frame unit. According to the height of the upper frame, adjust the 3D camera to the best shooting height, and complete the correction of the 3D camera shooting height.

[0084] Specifically, in the S500 step, after the single-layer frame is taken out, the height of the whole stack of materials will be reduced. Therefore, during the collection of the isolation paper point cloud and the frame point cloud, the height information returned by the point cloud is used to guide the robot to descend, so as to ensure that the shooting height of the 3D camera is within the best range. Therefore, during the grabbing of the frame, the stack position of the frame will be lowered, and accordingly the shooting height of the 3D camera will be lowered.

[0085] S600, repeat the steps S300-S500 until the whole stack of frames is taken out, and complete the unstacking of the stacked frames.

[0086] Since the two ends of the short frame have corner codes, corner code isolation paper is also arranged at the two ends of the short frame. Therefore, a step of grabbing the corner code isolation paper is needed before the short frame is grabbed. Therefore, before the single-layer frame is grabbed in step S400, step S700 of grabbing the corner code isolation paper is also needed, and the step of grabbing the corner code isolation paper in step S700 is as follows:

[0087] S701, the 3D camera moves above the corner code isolation paper to take a photo of the corner code isolation paper, collects the point cloud of the corner code isolation paper, and obtains the point cloud of the corner code isolation paper;

[0088] S702, the collected point cloud of the corner code isolation paper is preprocessed, the preprocessed point cloud of the corner code isolation paper is matched with a preset point cloud model of the corner code isolation paper, and the pose information of the point cloud of the corner code isolation paper is obtained;

[0089] S703, according to the pose information of the corner code isolation paper obtained in step S702, a corresponding grabbing clamp is selected, the corner code isolation paper is grabbed and put into the isolation paper box.

[0090] In step S702, the construction method of the corner code isolation paper point cloud model is consistent with the construction method of the frame isolation paper point cloud model, which will not be repeated here.

[0091] In order to complete the above-mentioned method for grabbing and unstacking the stacked frames based on 3D vision, a system for grabbing and unstacking the stacked frames based on 3D vision is also provided, which comprises:

[0092] The center photographing correction module photographs the four corners of the whole stack of frames with the 3D camera, determines the center of the whole stack of frames, and corrects the stack position of the whole stack of frames;

[0093] The point cloud acquisition module is used to acquire the point cloud data and depth information of the frame and the isolation paper;

[0094] The point cloud data processing module pre-processes the point cloud data and depth information of the point cloud acquisition module;

[0095] The point cloud matching module matches the preprocessed isolation paper point cloud with a preset isolation paper point cloud model to obtain the category information and pose information of the isolation paper, the category information of the isolation paper being frame isolation paper and corner code isolation paper; the preprocessed frame point cloud is matched with a preset frame point cloud model to obtain the type information and pose information of the frame, the distance between each two adjacent frames in the plurality of frames is calculated, and it is determined whether the distance meets the requirements;

[0096] The pose conversion module converts the pose information of the frame point cloud and the separator point cloud into the grasping pose information of the robot gripper by using the pose conversion matrix obtained by the hand-eye calibration model. The pose conversion matrix of the separator or the frame is output by the hand-eye calibration model, and the pose of the separator or the frame in the 3D camera can be converted into the pose in the grasping device module by the pose conversion matrix. The pose conversion matrix is obtained by calibrating the 3D camera by the eye-in-hand calibration method, which is a prior art and will not be described here.

[0097] The grasping device module receives the control instruction and the pose data output by the pose conversion module, and performs a grasping action to grasp the separator and the frame.

[0098] The photographing height correction module further obtains the height of the frame unit in the upper layer during photographing of the frame unit, adjusts the 3D camera to the optimal photographing height according to the height of the frame in the upper layer, and completes the correction of the photographing height of the 3D camera.

[0099] The above only describes some embodiments of the present application. Those skilled in the art can make several modifications and improvements without departing from the concept of the present application, which are all within the protection scope of the present application.

Claims

1. A method for capturing a single-layer border in a stacked border based on 3D vision, characterized in that: It includes the following steps: S1. Let m be the maximum number of borders that the border grabbing fixture can grab at one time, and M be the number of borders in a single layer of borders. Divide all the borders in a single layer of borders into n border units based on m, and n = [M / m] + 1. S2. Use a 3D camera to take a picture of the designated center area of ​​the first border unit, collect the point cloud data of m borders, and obtain the border point cloud. S3. Preprocess the bounding point cloud and then match it with the set bounding point cloud model to obtain the category information and pose information of the bounding. Calculate the spacing between each pair of adjacent boundings in the first bounding unit 1 to m. The category information of the bounding includes front bounding and back bounding. S4. Determine whether the spacing between any two adjacent borders in the 1 to m borders meets the set spacing range. If all meet the range, capture the 1 to m borders according to the category information and pose information obtained in step S3, complete the capture of the first border unit, and execute step S7; otherwise, execute steps S5 to S6. S5. Determine the border combinations that do not meet the set spacing range, and denote them as border combination i = {border A i , border B i}, that is, the spacing between adjacent border A i and border B i does not meet the set spacing range, i = 1, 2,..., p, 1 ≤ p < m - 1, where p is the number of border combinations determined to not meet the set spacing range; S6. Based on the category and pose information obtained in step S3, first capture the first bounding box A1, then capture bounding boxes B1 to A2, and so on, until bounding box B is captured. p ~The m-th border, completing the capture of the first border unit; S7. Follow steps S2 to S6 to complete the capture of all subsequent border units, thus completing the capture of a single-layer border.

2. A method for destacking and grasping stacking borders based on 3D vision, characterized in that: It includes the following steps: The S100 and 3D cameras take pictures of the top layer of the entire stack of frames to determine the overall height of the entire stack of frames. The S200 and 3D cameras take pictures of the four corners of the entire stack frame to determine the center of the entire stack frame and correct the stack position of the entire stack frame. S300, gripping of the edge release paper: The S301 3D camera moves above the frame isolation paper to take pictures of the frame isolation paper, and collects the point cloud of the frame isolation paper to obtain the point cloud of the frame isolation paper. S302. Preprocess the collected border isolation paper point cloud, match the preprocessed border isolation paper point cloud with the preset border isolation paper point cloud model, and obtain the pose information of the border isolation paper point cloud. S303. Based on the pose information of the frame isolation paper obtained in step S302, select the corresponding gripping fixture, grab the frame isolation paper and put it into the isolation paper material box. S400, Single-layer border capture: The single-layer border is captured according to the single-layer border capture method as described in claim 1. S500: During the upper border capture process, when taking a picture of the border unit, further obtain the height of the border layer, adjust the 3D camera to the optimal shooting height according to the height of the upper border, and complete the correction of the 3D camera shooting height. S600. Repeat steps S300 to S500 until all the edges of the stack are removed, thus completing the destacking of the stack edges.

3. The method for destacking and grasping stacking borders based on 3D vision as described in claim 2, characterized in that: Before step S400, which involves grabbing the single-layer border, there is step S700, which involves grabbing the corner code isolation paper. The steps for grabbing the corner code isolation paper in step S700 are as follows: The S701 3D camera moves above the corner code isolation paper to take pictures of the corner code isolation paper, and collects the point cloud of the corner code isolation paper to obtain the point cloud of the corner code isolation paper; S702. Preprocess the collected corner code isolation paper point cloud, match the preprocessed corner code isolation paper point cloud with the preset corner code isolation paper point cloud model, and obtain the pose information of the corner code isolation paper point cloud. S703. Based on the position information of the corner code isolation paper obtained in step S702, select the corresponding gripping fixture, grab the corner code isolation paper and put it into the isolation paper material box.

4. A stacking border destacking and gripping system based on 3D vision, characterized in that... The method for destacking and gripping the stacking edge as described in claim 2 includes: The centering photo correction module uses a 3D camera to take pictures of the four corners of the entire stack's frame, determine the center of the entire stack's frame, and correct the stack's position. The point cloud acquisition module is used to acquire point cloud data and depth information of the border and the isolation paper; The point cloud data processing module preprocesses the point cloud data and depth information from the point cloud acquisition module. The point cloud matching module matches the pre-processed isolation paper point cloud with a preset isolation paper point cloud model to obtain the category information and pose information of the isolation paper. The category information of the isolation paper is border isolation paper and corner code isolation paper. The module also matches the pre-processed border point cloud with a preset border point cloud model to obtain the type information and pose information of the border. It calculates the spacing between each pair of adjacent borders in multiple borders and determines whether the spacing meets the requirements. The pose conversion module uses the pose conversion matrix obtained from the hand-eye calibration model to convert the pose information of the bounding box point cloud and the isolation paper point cloud into the grasping pose information of the robot gripper. The grasping device module receives control commands and pose data output by the pose conversion module, and grasps the isolation paper and the frame in sequence. The image capture height correction module obtains the height of the border unit during the upper border capture process and adjusts the 3D camera to the optimal image capture height based on the upper border height, thus completing the correction of the 3D camera image capture height.

5. A destacking and gripping device for stacking borders based on 3D vision, characterized in that... The method for destacking and gripping the stacked edge as described in claim 2 includes a mounting plate disposed at the moving end of a robotic arm, and edge separation paper gripping fixtures, corner code separation paper gripping fixtures, and edge gripping fixtures respectively disposed on the mounting plate.

6. The photovoltaic frame destacking and gripping device based on 3D vision as described in claim 5, characterized in that: The edge separation paper gripper includes a first cylinder, a movable frame driven by the first cylinder to move up and down, and a plurality of first suction cups fixed at the lower end of the movable frame for picking up the edge separation paper.

7. The photovoltaic frame destacking and gripping device based on 3D vision as described in claim 5, characterized in that: The corner code isolation paper gripper includes a second cylinder and a pair of grippers driven by the second cylinder to open or clamp, used to grip the corner code isolation paper.

8. The photovoltaic frame destacking and gripping device based on 3D vision as described in claim 5, characterized in that: The frame gripping fixture includes a plurality of air passage blocks arranged at equal intervals along a first direction and mounted on the mounting plate, a plurality of second suction cups fixedly arranged at the lower end of the air passage blocks along a second direction, and a motor that drives the plurality of air passage blocks to simultaneously increase or decrease the spacing along the first direction. The first direction and the second direction are perpendicular to each other. When the frame gripping fixture picks up the frame, the first direction is the width direction of the frame and the second direction is the length direction of the frame.

Images