A control method and device for collaborative robot grabbing fruits and vegetables

Through identification and geometric matching algorithms to determine the grasping point and angle of the collaborative robot claw, the problem of unstable grasping caused by inaccurate estimation of fruit and vegetable poses in the prior art is solved, and the stable and labor-saving grasping effect of fruit and vegetable is achieved.

CN115302514BActive Publication Date: 2024-12-03JIANGHAN UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211027561.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-25
Publication Date
2024-12-03
Estimated Expiration
2042-08-25

AI Technical Summary

Technical Problem

The existing collaborative robots fail to accurately estimate the position of fruits and vegetables in fruit and vegetable sorting and picking, resulting in unstable and laborious gripping of the jaws, and unable to effectively grasp fruits and vegetables.

Method used

By obtaining the image to be identified, using the preset fruit and vegetable recognition model and matching algorithm for fruit and vegetable recognition and geometric matching, determine the grab point of the robot's jaw, and adjust the grab angle of the jaw according to the geometric matching figure to achieve stable fruit and vegetable grabbing Pick.

Benefits of technology

It improves the stability and efficiency of fruit and vegetable grabbing, and achieves a stable and labor-saving grasp of fruit and vegetable by robots.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115302514B_ABST
    Figure CN115302514B_ABST
Patent Text Reader

Abstract

The present application discloses a control method and device for a collaborative robot to grasp fruits and vegetables, the method comprising: obtaining a segmented image to be recognized; using a preset fruit and vegetable recognition model to recognize fruits and vegetables on the image to be recognized, and determining the fruits and vegetables to be grasped; using a preset matching algorithm to geometrically match the fruits and vegetables to be grasped, and obtaining a geometric matching graph; determining the grasping point of the robot gripper according to the geometric matching graph and the coordinate conversion relationship between the pre-acquired two-dimensional image space and the real space; adjusting the grasping angle of the robot gripper according to the grasping point, and controlling the robot to grasp the fruits and vegetables to be grasped. The present invention makes full use of the geometric information of the fruits and vegetables to be grasped, making the posture estimation of the fruits and vegetables more accurate, determining the most suitable grasping point according to the geometric matching graph, and grasping the fruits and vegetables according to the grasping point, so as to realize the robot's effortless and stable grasping of the fruits and vegetables.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural robots, and in particular to a control method and device for a collaborative robot to grasp fruits and vegetables. Background Art

[0002] With the rapid development of science and technology, collaborative robots are widely used in agricultural production. Collaborative robots are used to perform high-precision and high-repeat operations, solving the supply-demand imbalance problem of large-scale vegetable and fruit planting and insufficient labor in agricultural production. The addition of collaborative robots can improve the yield rate of fruit and vegetable picking and sorting, and also improve the automation and intelligence level of agricultural production.

[0003] In the application scenarios of fruit and vegetable sorting and picking, an important task of collaborative robots is to grab fruits and vegetables. However, in the prior art, the grabbing method of collaborative robots is generally to move the gripper to grab the fruits and vegetables that need to be grabbed after finding and identifying them, without estimating the position of the object. For example, when grabbing a cucumber, the best solution is to grab the cucumber along the center of the cucumber perpendicular to the long axis of the cucumber. This gripping method is the most stable and labor-saving. However, existing collaborative robots often grab the cucumber at any position as long as they identify the cucumber. This grabbing method is not only laborious, but also unstable, and often does not necessarily succeed.

[0004] Therefore, it is necessary to provide a control method for collaborative robots to grasp fruits and vegetables to solve the problem that existing collaborative robots fail to accurately estimate the posture of fruits and vegetables and effectively adjust the grippers during fruit and vegetable sorting and picking, resulting in the inability to grasp fruits and vegetables stably. Summary of the invention

[0005] In view of this, it is necessary to provide a control method and device for a collaborative robot to grasp fruits and vegetables, so as to solve the problem that the existing collaborative robots fail to accurately estimate the posture of objects and effectively adjust the grippers, resulting in the inability to grasp fruits and vegetables stably.

[0006] In order to solve the above problems, the present invention provides a control method for a collaborative robot to grab fruits and vegetables, comprising:

[0007] Obtain the segmented image to be recognized;

[0008] Using a preset fruit and vegetable recognition model to perform fruit and vegetable recognition on the image to be recognized, and determine the fruits and vegetables to be grasped;

[0009] Using a preset matching algorithm to geometrically match the fruits and vegetables to be grasped, to obtain a geometric matching graph;

[0010] Determine the gripping point of the robot gripper according to the geometric matching figure and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space;

[0011] The gripping angle of the robot gripper is adjusted according to the gripping point, and the robot is controlled to grip the fruits and vegetables to be gripped.

[0012] Furthermore, a preset matching algorithm is used to perform geometric matching on the fruits and vegetables to be grasped to obtain a geometric matching graph, including:

[0013] Using a minimum frame symmetrical geometric figure algorithm, shape matching is performed on the fruits and vegetables to be grasped through different geometric figures to obtain different geometric matching shapes;

[0014] Among the different geometric matching shapes, the one with the largest area occupied by the fruits and vegetables to be grasped is selected as the geometric matching shape.

[0015] Furthermore, the minimum frame symmetrical geometric figure algorithm is used to perform shape matching on the fruits and vegetables to be grasped through different geometric figures to obtain different geometric matching shapes, including:

[0016] Use the rotating caliper algorithm to frame the minimum matching rectangle of the fruits and vegetables to be grasped;

[0017] The Hatchyan algorithm is used to define the minimum matching ellipse of the fruits and vegetables to be grasped;

[0018] The minimum circle covering algorithm is used to define the minimum matching circle of the fruits and vegetables to be grasped.

[0019] Further, according to the geometric matching graph and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space, the grasping point of the robot gripper is determined, including:

[0020] When the geometric matching figure is a rectangle or an ellipse, the two end points of the short axis of the geometric matching figure are used as the actual positions corresponding to the boundaries of the fruits and vegetables to be grasped;

[0021] When the geometric matching figure is a circle, two end points of the diameter of the geometric matching image are used as grabbing points at actual positions corresponding to the boundaries of the fruits and vegetables to be grabbed.

[0022] Further, the gripper of the robot is a two-finger gripper;

[0023] Adjusting the gripping angle of the robot gripper according to the gripping point includes:

[0024] Moving the gripper of the robot so that the axis of the gripper of the robot passing through the center point of the geometric matching figure is perpendicular to the plane where the geometric matching figure is located;

[0025] Determine the offset angle formed by the line connecting the grasping points and the line connecting the midpoints of the two gripping fingers of the robot gripper;

[0026] The angle of the clamping jaw is adjusted according to the offset angle.

[0027] Further, adjusting the angle of the clamping jaw according to the offset angle includes:

[0028] When the offset angle is an acute angle, rotating the clamp counterclockwise by the offset angle;

[0029] When the offset angle is an obtuse angle, the clamping jaw is rotated clockwise by a complementary angle of the offset angle.

[0030] Furthermore, controlling the robot to grab the fruits and vegetables to be grabbed also includes:

[0031] The extension width of the clamping jaw is determined according to the width of the line connecting the gripping points.

[0032] Furthermore, a torque sensor is provided on the robot gripper;

[0033] Controlling the robot to grab the fruits and vegetables to be grabbed also includes:

[0034] The torque sensor is used to obtain the magnitude of the friction force between the robot gripper and the fruits and vegetables, and the gripping force of the gripper is determined according to the friction force.

[0035] Furthermore, the preset fruit and vegetable recognition model is constructed based on a convolutional neural network.

[0036] The present invention also provides a device for a collaborative robot to grab fruits and vegetables, comprising:

[0037] An image acquisition module is used to acquire the segmented image to be recognized;

[0038] A fruit and vegetable recognition module, used to perform geometric matching on the fruits and vegetables to be grasped using a preset matching algorithm to obtain a geometric matching graph;

[0039] A grasping point determination module is used to determine the grasping point of the robot gripper according to the geometric matching figure and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space;

[0040] The grasping control module is used to adjust the grasping angle of the robot gripper according to the grasping point, and control the robot to grasp the fruits and vegetables to be grasped.

[0041] Compared with the prior art, the beneficial effects of the present invention include: first, obtaining a picture to be identified, and identifying the fruits and vegetables to be grasped in the image through a fruit and vegetable recognition model; second, geometrically matching the fruits and vegetables to be grasped to determine the grasping points; finally, adjusting the grasping angle of the robot according to the grasping points to grasp the fruits and vegetables. The present invention identifies the fruits and vegetables to be grasped in the image through a fruit and vegetable recognition model, and geometrically matches the fruits and vegetables, making full use of the geometric information of the fruits and vegetables to be grasped, making the posture estimation of the fruits and vegetables more accurate; determining the most suitable grasping points according to the geometric matching graphics, and grasping the fruits and vegetables according to the grasping points, so that the robot can grasp the fruits and vegetables in a labor-saving and stable manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of an embodiment of a control method for a collaborative robot to grasp fruits and vegetables provided by the present invention;

[0043] FIG2( a) is a schematic diagram of an embodiment of a detection image acquired by a robot wrist camera provided by the present invention;

[0044] FIG2( b ) is a schematic diagram of an embodiment of a segmented image provided by the present invention;

[0045] FIG2( c ) is a schematic diagram of an embodiment of the present invention for shape matching of fruits and vegetables using a circle;

[0046] FIG2( d ) is a schematic diagram of an embodiment of shape matching of fruits and vegetables using rectangles provided by the present invention;

[0047] FIG2( e ) is a schematic diagram of an embodiment of shape matching of fruits and vegetables using an ellipse provided by the present invention;

[0048] Figure 3 A schematic diagram of an embodiment of determining the gripping point of a robot gripper according to a geometric matching pattern provided by the present invention;

[0049] Figure 4 This is a structural schematic diagram of an embodiment of a collaborative robot device for grasping fruits and vegetables provided by the present invention. DETAILED DESCRIPTION

[0050] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0051] Before describing the embodiments, the inventive concept of the present application is first explained:

[0052] In the application scenarios of fruit and vegetable sorting and picking, the existing collaborative robots generally grasp fruits and vegetables by finding and identifying the fruits and vegetables to be grasped and then moving the gripper to grasp them. However, they do not analyze the position of the objects and do not select the most suitable grasping points. Therefore, the success rate of grasping is not high.

[0053] Since fruits and vegetables often have symmetrical geometric shapes, the present application uses symmetrical geometric figures to match the fruits and vegetables to be grasped, and grasps the fruits and vegetables along the short axis direction of the geometric matching figures of the fruits and vegetables to achieve the most stable and labor-saving grasping effect. Based on the geometric shape information of fruits and vegetables, the present invention accurately estimates the position and posture of fruits and vegetables to improve the robot's grasping stability of fruits and vegetables and achieve stable grasping of fruits and vegetables by the robot.

[0054] The embodiment of the present invention provides a control method for a collaborative robot to grasp fruits and vegetables. Figure 1 A flow chart of an embodiment of a control method for a collaborative robot to grasp fruits and vegetables provided by the present invention includes:

[0055] Step S101: obtaining the segmented image to be recognized;

[0056] Step S102: using a preset fruit and vegetable recognition model to perform fruit and vegetable recognition on the image to be recognized, and determining the fruits and vegetables to be grasped;

[0057] Step S103: using a preset matching algorithm to perform geometric matching on the fruits and vegetables to be grasped, to obtain a geometric matching graph;

[0058] Step S104: determining the gripping point of the robot gripper according to the geometric matching pattern and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space;

[0059] Step S105: adjusting the gripping angle of the robot gripper according to the gripping point, and controlling the robot to grip the fruits and vegetables to be gripped.

[0060] The control method for the collaborative robot to grasp fruits and vegetables provided in this embodiment first obtains the image to be identified, and recognizes the fruits and vegetables to be grasped in the image through the fruit and vegetable recognition model; secondly, geometric matching is performed on the fruits and vegetables to be grasped to determine the grasping points; finally, the grasping angle of the robot is adjusted according to the grasping points to grasp the fruits and vegetables. This embodiment recognizes the fruits and vegetables to be grasped through the fruit and vegetable recognition model, and geometrically matches the fruits and vegetables, making full use of the geometric information of the fruits and vegetables to be grasped, making the position and posture estimation of the fruits and vegetables more accurate; the most suitable grasping points are determined according to the geometric matching graphics, and the fruits and vegetables are grasped according to the grasping points, so that the robot can grasp the fruits and vegetables in a labor-saving and stable manner.

[0061] As a specific embodiment, in step S101, the segmented image to be identified is acquired through the robot's wrist camera, specifically by controlling the robotic arm of the collaborative robot to move above the operating table and parallel to the operating table, using the wrist camera provided on the robotic arm to acquire the detection image of the operating table in real time, and segmenting the detection image to obtain the segmented image to be identified.

[0062] As a specific embodiment, the preset image segmentation method is the Mean-Shift algorithm, and the object segmentation is performed on the detection image by the Mean-Shift algorithm to obtain a segmented image.

[0063] In order to completely detect the object on the operating table, as a specific embodiment, when the segmented object is located in the center of the image, the robot arm stops moving and the current segmented image is stored for subsequent object recognition.

[0064] As a preferred embodiment, in step S102, the preset fruit and vegetable recognition model is constructed based on a convolutional neural network, the fruit and vegetable recognition model is obtained by the trained convolutional neural network, and the fruits and vegetables in the segmented image are recognized by the fruit and vegetable recognition model.

[0065] As a preferred embodiment, in step S103, a preset matching algorithm is used to perform geometric matching on the fruits and vegetables to be grasped to obtain a geometric matching graph, including:

[0066] Using a minimum frame symmetrical geometric figure algorithm, shape matching is performed on the fruits and vegetables to be grasped through different geometric figures to obtain different geometric matching shapes;

[0067] Among the different geometric matching shapes, the one with the largest area occupied by the fruits and vegetables to be grasped is selected as the geometric matching shape.

[0068] As a preferred embodiment, the minimum frame symmetrical geometric figure algorithm is used to perform shape matching on the fruits and vegetables to be grasped through different geometric figures to obtain different geometric matching shapes, including:

[0069] Use the rotating caliper algorithm to frame the minimum matching rectangle of the fruits and vegetables to be grasped;

[0070] The Hatchyan algorithm is used to define the minimum matching ellipse of the fruits and vegetables to be grasped;

[0071] The minimum circle covering algorithm is used to define the minimum matching circle of the fruits and vegetables to be grasped.

[0072] Below through Figure 2(a) to Figure 2(e) The above object segmentation and shape matching process is explained.

[0073] As shown in Figure 2(a), the detection image obtained by the robot wrist camera is shown in Figure 2(a); the objects in the image are segmented using the Mean-Shift algorithm to obtain Figure 2(b).

[0074] The shapes of the objects are matched by different geometric figures. FIG2(c), FIG2(d) and FIG2(e) respectively show the geometric matching process of using a circle, a rectangle and an ellipse to perform the minimum frame on the fruits and vegetables. The geometric shape whose area of ​​the fruits and vegetables accounts for the largest area of ​​the matching shape is used as the geometric matching figure. In this embodiment, after the area ratio analysis, the ellipse is selected as the geometric matching figure of the fruits and vegetables.

[0075] As a preferred embodiment, in step S104, the grasping point of the robot gripper is determined according to the geometric matching graph and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space, including:

[0076] When the geometric matching figure is a rectangle or an ellipse, the two end points of the short axis of the geometric matching figure are used as the actual positions corresponding to the boundaries of the fruits and vegetables to be grasped;

[0077] When the geometric matching figure is a circle, two end points of the diameter of the geometric matching image are used as grabbing points at actual positions corresponding to the boundaries of the fruits and vegetables to be grabbed.

[0078] As a specific embodiment, the coordinate conversion relationship between the two-dimensional image space and the real space is obtained by calibrating the robot's wrist camera. Since the camera can only obtain a two-dimensional image before it is calibrated, in order to determine the three-dimensional position relationship of the object photographed by the camera in the real scene, a camera calibration method commonly used by technicians in this field, such as the chessboard calibration method, is used to calibrate the robot's wrist camera, thereby obtaining the coordinate conversion relationship between the two-dimensional image space and the real space.

[0079] As a specific embodiment, when the geometric matching figure is a rectangle or an ellipse, the actual positions of the major axis and the minor axis of the geometric matching figure on the fruit and vegetable boundary are determined, and the actual physical position corresponding to the minor axis of the geometric matching figure on the fruit and vegetable boundary is used as the grasping point. When the geometric matching figure is a circle, the actual physical positions corresponding to the two end points of any diameter of the circle on the fruit and vegetable boundary are determined, and the actual positions corresponding to the end points of the diameter are used as the grasping points.

[0080] As a preferred embodiment, the gripper of the robot is a two-finger gripper;

[0081] In step S105, adjusting the gripping angle of the robot gripper according to the gripping point includes:

[0082] Moving the gripper of the robot so that the axis of the gripper of the robot passing through the center point of the geometric matching figure is perpendicular to the plane where the geometric matching figure is located;

[0083] Determine the offset angle formed by the line connecting the grasping points and the line connecting the midpoints of the two gripping fingers of the robot gripper;

[0084] The angle of the clamping jaw is adjusted according to the offset angle.

[0085] As a preferred embodiment, adjusting the angle of the clamping jaw according to the offset angle includes:

[0086] When the offset angle is an acute angle, rotating the clamp counterclockwise by the offset angle;

[0087] When the offset angle is an obtuse angle, the clamping jaw is rotated clockwise by a complementary angle of the offset angle.

[0088] As a preferred embodiment, controlling the robot to grab the fruits and vegetables to be grabbed also includes:

[0089] The extension width of the clamping jaw is determined according to the width of the line connecting the gripping points.

[0090] As a specific embodiment, if the extension width of the clamp is larger than the short axis of the geometric matching figure corresponding to the fruit and vegetable, the clamp can be moved downward and tightened to the grasping point.

[0091] Combine the following Figure 3 The above technical solution is illustrated through the robot's grasping process of cucumbers.

[0092] As a specific example, Figure 3 As shown, a rectangle is used as a geometric matching pattern for cucumbers, and the robot arm moves along the long axis of the rectangle. When the robot detects that the cucumber is located in the center of the image, the robot arm stops moving, so that the line connecting the center of the wrist camera and the center of the geometric matching pattern is perpendicular to the operating table, and is ready to pick up the fruits and vegetables. Determine the offset angle formed by the line connecting the grasping points and the line connecting the midpoints of the two gripping fingers of the robot gripper, and adjust the gripping angle of the gripper according to the offset angle. Adjust the extended width of the gripper to be larger than the short axis of the rectangle, and move it down to the grasping point to grab the cucumber.

[0093] As a preferred embodiment, a torque sensor is provided on the robot gripper;

[0094] Controlling the robot to grab the fruits and vegetables to be grabbed also includes:

[0095] The torque sensor is used to obtain the magnitude of the friction force between the robot gripper and the fruits and vegetables, and the gripping force of the gripper is determined according to the friction force.

[0096] In order to effectively grip all fruits and vegetables on the operating table, as a specific embodiment, the moving path of the collaborative robot arm is preset, and the robot identifies and grips the fruits and vegetables to be gripped along the preset moving path.

[0097] This embodiment also provides a device for a collaborative robot to grab fruits and vegetables, and its structural block diagram is as follows: Figure 4 As shown, the device 400 for grabbing fruits and vegetables by a collaborative robot includes:

[0098] An image acquisition module 401 is used to acquire the segmented image to be recognized;

[0099] The fruit and vegetable recognition module 402 is used to use a preset fruit and vegetable recognition model to perform fruit and vegetable recognition on the image to be recognized, and determine the fruit and vegetable to be grabbed;

[0100] The geometric matching module 403 is used to perform geometric matching on the fruits and vegetables to be grasped by using a preset matching algorithm to obtain a geometric matching graph;

[0101] A grasping point determination module 404 is used to determine the grasping point of the robot gripper according to the geometric matching pattern and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space;

[0102] The grasping control module 405 is used to adjust the grasping angle of the robot gripper according to the grasping point, and control the robot to grasp the fruits and vegetables to be grasped.

[0103] The present invention discloses a control method and device for a collaborative robot to grasp fruits and vegetables. First, a calibrated wrist camera of the robot is used to obtain a detection image; second, objects in the image are segmented, and the fruits and vegetables to be grasped in the image are identified by a fruit and vegetable recognition model; third, geometric matching is performed on the fruits and vegetables to be grasped to determine the grasping points; finally, the grasping angle of the robot is adjusted according to the grasping points to grasp the fruits and vegetables.

[0104] The present invention obtains more detailed object images through a calibrated wrist camera, identifies the fruits and vegetables to be grasped through image segmentation and fruit and vegetable recognition models, and performs geometric shape matching on the fruits and vegetables, making full use of the geometric information of the fruits and vegetables to be grasped, making the posture estimation of the fruits and vegetables more accurate; determines the most suitable grasping points according to the geometric matching graphics, and grasps the fruits and vegetables according to the grasping points, thereby realizing the robot's labor-saving and stable grasping of fruits and vegetables.

[0105] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A control method for a collaborative robot to grab fruits and vegetables, characterized in that: include: Acquire the segmented image to be identified; use a preset fruit and vegetable identification model to identify fruits and vegetables in the image to be identified, and determine the fruits and vegetables to be grasped; Using a preset matching algorithm to geometrically match the fruits and vegetables to be grasped, to obtain a geometric matching graph; Determine the gripping point of the robot gripper according to the geometric matching figure and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space; Adjusting the gripping angle of the robot gripper according to the gripping point, and controlling the robot to grip the fruits and vegetables to be gripped; The fruits and vegetables to be grasped are geometrically matched using a preset matching algorithm to obtain a geometric matching graph, including: Using a minimum frame symmetrical geometric figure algorithm, shape matching is performed on the fruits and vegetables to be grasped through different geometric figures to obtain different geometric matching shapes; Selecting, among the different geometric matching shapes, the one with the largest area occupied by the fruits and vegetables to be grasped as the geometric matching shape; Using the minimum frame symmetrical geometric figure algorithm, the fruits and vegetables to be grasped are matched in shape through different geometric figures to obtain different geometric matching shapes, including: Use the rotating caliper algorithm to frame the minimum matching rectangle of the fruits and vegetables to be grasped; The Hatchyan algorithm is used to define the minimum matching ellipse of the fruits and vegetables to be grasped; The minimum circle covering algorithm is used to define the minimum matching circle of the fruits and vegetables to be grasped.

2. The control method for a collaborative robot to grasp fruits and vegetables according to claim 1, characterized in that: Determining the grasping point of the robot gripper according to the geometric matching graph and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space includes: When the geometric matching figure is a rectangle or an ellipse, the two end points of the short axis of the geometric matching figure are used as the actual positions corresponding to the boundaries of the fruits and vegetables to be grasped; When the geometric matching figure is a circle, two end points of the diameter of the geometric matching image are used as grabbing points at actual positions corresponding to the boundaries of the fruits and vegetables to be grabbed.

3. The control method for a collaborative robot to grasp fruits and vegetables according to claim 1, characterized in that: The gripper of the robot is a two-finger gripper; Adjusting the gripping angle of the robot gripper according to the gripping point includes: Moving the gripper of the robot so that the axis of the gripper of the robot passing through the center point of the geometric matching figure is perpendicular to the plane where the geometric matching figure is located; Determine the offset angle formed by the line connecting the grasping points and the line connecting the midpoints of the two gripping fingers of the robot gripper; The angle of the clamping jaw is adjusted according to the offset angle.

4. The control method for a collaborative robot to grasp fruits and vegetables according to claim 3, characterized in that: The angle of the clamping jaw is adjusted according to the offset angle, comprising: When the offset angle is an acute angle, rotating the clamp counterclockwise by the offset angle; When the offset angle is an obtuse angle, the clamping jaw is rotated clockwise by a complementary angle of the offset angle.

5. The control method for a collaborative robot to grasp fruits and vegetables according to claim 3, characterized in that: Controlling the robot to grab the fruits and vegetables to be grabbed also includes: The extension width of the clamping jaw is determined according to the width of the line connecting the gripping points.

6. The control method for a collaborative robot to grasp fruits and vegetables according to claim 1, characterized in that: A torque sensor is provided on the robot gripper; Controlling the robot to grab the fruits and vegetables to be grabbed also includes: The torque sensor is used to obtain the magnitude of the friction force between the robot gripper and the fruits and vegetables, and the gripping force of the gripper is determined according to the friction force.

7. The control method for a collaborative robot to grasp fruits and vegetables according to claim 1, characterized in that: The preset fruit and vegetable recognition model is constructed based on a convolutional neural network.

8. A control device for a collaborative robot to grab fruits and vegetables, characterized in that: An image acquisition module is used to acquire the segmented image to be recognized; A fruit and vegetable recognition module, used to use a preset fruit and vegetable recognition model to perform fruit and vegetable recognition on the image to be recognized, and determine the fruit and vegetable to be grabbed; A geometric matching module, used to perform geometric matching on the fruits and vegetables to be grasped using a preset matching algorithm to obtain a geometric matching graph; A grasping point determination module is used to determine the grasping point of the robot gripper according to the geometric matching figure and the coordinate transformation relationship between the pre-acquired two-dimensional image space and the real space; A grasping control module, used to adjust the grasping angle of the robot gripper according to the grasping point, and control the robot to grasp the fruits and vegetables to be grasped; The fruits and vegetables to be grasped are geometrically matched using a preset matching algorithm to obtain a geometric matching graph, including: Using a minimum frame symmetrical geometric figure algorithm, shape matching is performed on the fruits and vegetables to be grasped through different geometric figures to obtain different geometric matching shapes; Selecting, among the different geometric matching shapes, the one with the largest area occupied by the fruits and vegetables to be grasped as the geometric matching shape; Using the minimum frame symmetrical geometric figure algorithm, the fruits and vegetables to be grasped are matched in shape through different geometric figures to obtain different geometric matching shapes, including: Use the rotating caliper algorithm to frame the minimum matching rectangle of the fruits and vegetables to be grasped; The Hatchyan algorithm is used to define the minimum matching ellipse of the fruits and vegetables to be grasped; The minimum circle covering algorithm is used to define the minimum matching circle of the fruits and vegetables to be grasped.