CN110509281A - Device and method for pose recognition and grasping based on binocular vision - Google Patents

Abstract

The invention discloses a device and method for binocular vision-based pose recognition and grasping. It includes a workpiece transmission module, an image acquisition module and a target workpiece grabbing module. The workpiece transmission module transmits the target workpiece to the bottom of the binocular camera through the conveyor belt. When the proximity sensor installed under the binocular vision camera bracket receives the workpiece arrival signal, the conveyor belt stops working. It is convenient for left and right industrial cameras to collect images; the image acquisition module collects images and transmits them to the computer, and obtains the pose information of the target workpiece after processing through the image preprocessing algorithm and pose recognition algorithm, and then processes the pose information, trajectory planning, and the data Transfer to the grabbing device; the manipulator in the grabbing device grabs the workpiece and places it in the designated area. The present invention can detect the position and posture information of the target workpiece and perform trajectory planning, and use the grasping device to grasp the target workpiece, thereby improving the efficiency.

Description

Device and method for pose recognition and grasping based on binocular vision

technical field

The invention belongs to the field of machine vision, and in particular relates to a binocular vision-based pose recognition and grasping device and method.

Background technique

Binocular stereo vision is an important form of machine vision. It is based on the principle of parallax and uses imaging equipment to obtain two images of the measured object from different positions, and obtains the three-dimensional geometry of the object by calculating the position deviation between the corresponding points of the image. method of information.

With the rapid development of information technology and robot technology, the content covered by robot control technology is becoming more and more abundant. As a new type of digital control equipment, industrial robots have attracted widespread attention since they came out. After continuous exploration by experts, scholars and researchers at home and abroad, their functions have been gradually improved. Among them, binocular stereo vision technology is widely used in object recognition, virtual reality, industrial inspection and other fields. Binocular stereo vision technology can flexibly obtain stereo information of objects under various conditions, and has incomparable advantages over monocular vision.

In the industrial production line, the traditional feeding equipment is a vibrating plate, but for weak impact tough materials such as soft magnets and ceramic sheets, the vibrating plate cannot be used for feeding, otherwise the material will be broken and cracked. In the application of industrial manipulators, the general action of manipulators based on monocular vision is to use the "grab-place" action, using monocular vision to perform two-dimensional measurement of parts, and obtain the rotation angle and position of parts through fixed parameters. attitude to achieve the grabbing of parts. In the scenario of using binocular vision industrial manipulators, the grasping of parts and products by the manipulator only stays at the level of "positioning-grabbing-placement", without considering the initial and final posture of the product. Therefore, on the basis of binocular vision, the research on the pose recognition of the workpiece and the trajectory planning of the manipulator has great positive significance for the improvement of industrial production efficiency.

Contents of the invention

In order to overcome the deficiencies of the prior art, the present invention proposes a binocular vision-based pose recognition and grasping device and method, which can complete the pose recognition and grasping of workpieces that cannot be loaded on the vibrating plate, including Through the binocular vision system and related image processing and 3D reconstruction algorithm, the relevant image algorithm is used to recognize the attitude of the workpiece, and the motion trajectory planning of the manipulator is used to realize the recognition and grasping of the workpiece.

The technical scheme that the present invention adopts is as follows:

1. A device for pose recognition and grasping based on binocular vision

Including workpiece transfer module, image acquisition module and target workpiece grabbing module;

The workpiece transmission module includes a belt conveyor, a belt conveyor motor drive and a proximity sensor. A conveyor belt for transmitting target workpieces is installed on the belt conveyor, and a proximity sensor is installed on the belt conveyor frame at one side of the conveyor belt.

The image acquisition module includes an industrial camera, a camera bracket, a strip light source and a binocular camera fixture. The camera bracket is fixed on the belt conveyor frame on both sides of the conveyor belt, and a height-adjustable binocular camera is installed between the two sides of the camera bracket. Camera fixture, two CCD industrial cameras are slidingly installed on the guide rail of the binocular camera fixture. The two industrial cameras are used to shoot the target workpiece passing the proximity sensor. The two industrial cameras are the left camera and the right camera respectively; the bottom of both sides of the camera bracket Two strip light sources arranged symmetrically are installed, and the strip light sources are used to provide illumination during image acquisition; the proximity sensor is arranged close to the camera bracket.

The target workpiece grabbing module includes a four-axis manipulator and a control cabinet. The control cabinet is located on the side of the belt conveyor and its height is consistent with the height of the conveyor belt. The height of the control cabinet and the conveyor belt is consistent so that the height of the target workpiece is consistent with the origin of the coordinate system of the manipulator, which is convenient for coordinate conversion. ; The four-axis manipulator is fixed on the control cabinet through its base.

2. A method of pose recognition and grasping based on binocular vision using the above-mentioned device

Include the following steps:

Step 1: Adjust the height of the industrial camera, the shooting angle and the distance between the two industrial cameras so that the target workpiece is within the focal length range of the industrial camera;

The binocular camera fixture moves up and down along the vertical rods on both sides of the camera bracket to adjust the height of the industrial camera, and the two industrial cameras move along the guide rails of the binocular camera fixture to realize the distance between the two industrial cameras. The installation angle of the industrial camera realizes the adjustment of the shooting angle of the industrial camera.

Step 2: According to the Zhang Zhengyou calibration method, perform double-target calibration on the left and right cameras to obtain the internal and external parameter matrix of the camera, and perform stereo calibration on the left and right cameras according to the parameter matrix to obtain the correction map;

Step 3: Select the hand-to-eye (Eye-To-Hand) hand-eye calibration mode for the four-axis manipulator and industrial camera to perform hand-eye calibration, and obtain the pose transformation relationship between the camera coordinate system and the manipulator coordinate system;

Among them, the camera coordinate system is established by the left camera of the industrial camera as the origin, and the manipulator coordinate system is established by the four-axis manipulator base as the origin;

Step 4: The target workpiece to be grabbed is conveyed to the position below the binocular camera through the belt conveyor, and when the proximity sensor receives the signal of the target workpiece, the belt conveyor motor drive device stops the work of the belt conveyor;

Step 5: The left and right cameras capture images of the target workpiece passing through the proximity sensor, and transmit the captured image of the target workpiece to the computer;

Step 6: Use the correction map obtained in step 1 to correct the target workpiece images collected by the left and right cameras, and then perform image segmentation, grayscale transformation, median filtering and edge detection preprocessing on the images in sequence;

Step 7: Use the NCC similarity measurement algorithm to perform stereo matching on the image preprocessed in step 6, find matching points through the pyramid model of binocular stereo vision, and form a depth map from the point cloud image formed by gathering the coordinates of all matching points. Calculate the average depth of the depth map to obtain the three-dimensional coordinates of the target workpiece in the camera coordinate system;

Step 8: Estimating the attitude of the target workpiece based on Halcon to obtain the attitude information Q of the target workpiece;

Step 9: Know the initial position and initial attitude of the end effector of the four-axis manipulator (that is, the three-dimensional coordinates and attitude information of the end effector in the initial state are known), use the Cartesian position control mode for trajectory planning, and the Cartesian position control mode Trajectory planning is performed according to the azimuth angle of the target workpiece relative to the end effector and the three-dimensional coordinates of the target workpiece. The end effector of the four-axis manipulator moves from the initial position to the vicinity of the target workpiece according to the planned trajectory, and the end effector always points to the target workpiece during the movement process. The target workpiece can be grabbed more efficiently;

Step 10: When the buffer distance between the end effector of the four-axis manipulator and the target workpiece is d, adjust the posture information of the end effector of the four-axis manipulator, and then approach the target workpiece in a gradual manner to complete the grasping;

Step 11: When the four-axis manipulator grabs the target workpiece, the four-axis manipulator first grabs the target workpiece and moves up the set distance H along the Z axis of the manipulator coordinate system, and then the four-axis manipulator controls the end effector to move to the initial position. Then, the attitude information Q and the attitude information of the end effector in the initial state are solved by inverse motion, and then the end effector of the manipulator is adjusted back to the initial attitude of the end effector. At this time, the target workpiece completes the transfer process from the initial attitude to the intermediate attitude;

Step 12: The four-axis manipulator continues to grab the target workpiece to the target stacking platform, and the target workpiece is in the final posture, and the positioning-grabbing-placement process of the target workpiece is completed.

Described step eight specifically is: first utilize SolidWorks to carry out 3D image drawing to target workpiece, and the target workpiece pose in 3D image is ideal pose; Import the 3D image of workpiece into Halcon and read, use create_model_3d( ) operator to generate a 3D shape matching template, use the operator find_shape_model_3d() to match the preprocessed image obtained in step 5 with the 3D shape matching template, and obtain the attitude information Q of the target workpiece relative to the manipulator coordinate system.

The azimuth angle of the target workpiece relative to the end effector in the step 9 is obtained by the following calculation: first, the three-dimensional coordinates of the target workpiece in the camera coordinate system are converted into the position of the target workpiece in the manipulator coordinate system through the pose conversion relationship in step 2 Three-dimensional coordinates, calculate the azimuth angle of the target workpiece relative to the end effector according to the three-dimensional coordinates of the target workpiece and the end effector in the manipulator coordinate system.

Adjust the attitude information of the four-axis manipulator end effector in step 10 according to the parameters m, n, and q. The calculation methods of the parameters m, n, and q are as follows:

In the manipulator coordinate system, set the counterclockwise rotation angles of the end effector of the four-axis manipulator around the x-axis, y-axis, and z-axis as m, n, and q, where m, n, and q are the up-down, left-right, and Parameters of the rotational movement;

Since the end effector always points to the object during the grasping process, the parameter m = α;

Among them, m is the rotation angle of the end effector around the x-axis of the manipulator coordinate system (the starting position of the rotation is the z-axis), and α is the azimuth angle of the target workpiece relative to the end effector along the x-axis;

The parameters n and q are calculated according to the attitude information Q of the target workpiece, which includes the Euler angles θ _x , θ _y , and θ _z of the target workpiece rotating around the x, y, and z axes of the manipulator coordinate system, so that The rotation angle n of the end effector around the y-axis of the manipulator coordinate system is θ _y , and the rotation angle q around the z-axis of the manipulator coordinate system is θ _z :

That is, n = θ _y

q ₌ θz

The workpiece coordinate system is the coordinate system established with the target workpiece as the origin.

The attitude information when the end effector of the four-axis manipulator 1 grabs the target workpiece 10 is the same as the attitude information Q of the target workpiece 10 in step eight.

The posture information Q of the target workpiece is transmitted to the control system of the four-axis manipulator 1 based on the TCP/IP protocol SOCKET communication, so that the four-axis manipulator completes the grasping of the target workpiece.

Beneficial effects of the present invention:

The present invention can complete the pose recognition of workpieces that cannot be loaded on the vibrating plate, and use the binocular vision system, related image processing algorithms and three-dimensional reconstruction algorithms to perform pose recognition of the target workpiece to obtain pose information, and the pose information includes the position and pose of the workpiece. Attitude parameters and position distance, and then process the position and attitude information, trajectory planning, and transmit the data to the grabbing device. Through trajectory planning, the target workpiece is captured with the optimal posture and placed on the target stacking platform, which improves to a certain extent. Improve the efficiency of grabbing the target workpiece.

Description of drawings

Fig. 1 is the overall structural representation of device of the present invention;

Fig. 2 is the flow chart of work of workpiece transmission device among the present invention;

Fig. 3 is the flowchart of binocular vision system work among the present invention;

Fig. 4 is the work flowchart of the grasping device of target workpiece among the present invention;

FIG. 5 is a schematic diagram of a camera coordinate system.

Fig. 6 is a schematic diagram of a workpiece coordinate system.

Fig. 7 is a schematic diagram of the manipulator coordinate system.

Figure 8 is a schematic diagram of the position of the end effector of the four-axis manipulator in the manipulator coordinate system

In the figure: 1. Four-axis manipulator, 2. Industrial camera, 3. Camera bracket, 4. Belt conveyor motor drive device, 5. Strip light source, 6. Target workpiece stacking platform, 7. Control cabinet, 8. Proximity Sensor, 9. Belt conveyor, 10. Target workpiece, 11. Binocular camera fixture.

Detailed ways

The present invention will be further described below in conjunction with drawings and embodiments.

As shown in Figure 1, the present invention comprises workpiece transmission module, image acquisition module and target workpiece grasping module; Workpiece transmission module comprises belt conveyor 9, belt conveyor motor driver 4 and proximity sensor 8, and belt conveyor 9 is equipped with a conveyor belt for transmitting the target workpiece 10, and a proximity sensor 8 is installed on the belt conveyor 9 frame at one side of the conveyor belt; the image acquisition module includes an industrial camera 2, a camera bracket 3, a strip light source 5 and a binocular Camera fixture 11, camera support 3 is fixed on the belt conveyor 9 frame that is positioned at both sides of conveyer belt, and height-adjustable binocular camera fixture 11 is installed between the two sides of camera support 3, on the guide rail of binocular camera fixture 11 Two CCD industrial cameras 2 are slidingly installed, and the two industrial cameras 2 are used to photograph the target workpiece 10 passing the proximity sensor 8. The two industrial cameras 2 are respectively the left camera and the right camera; the bottom of the camera bracket 3 is installed with symmetrical arrangements Two strip light sources 5, the strip light source 5 is used to provide illumination during image acquisition; the proximity sensor 8 is arranged close to the camera support 3; the target workpiece grasping module includes a four-axis manipulator 1 and a control cabinet 7, and the control cabinet 7 is located in the belt type One side of the conveyor 9 and the height is consistent with the height of the conveyor belt; the four-axis manipulator 1 is fixed on the control cabinet through its base.

As shown in Figure 2, the workpiece conveying device transmits the target workpiece 10 to the vicinity of the proximity sensor 8, and the proximity sensor 8 detects the approach of the target workpiece 10, generates a signal, and transmits the signal to the belt conveyor motor drive device 4 to stop the conveyor belt motor work, the brake works, and finally the conveyor belt is stopped, and the workpiece is now near the bottom of the left and right CCD industrial cameras 2, which can facilitate the work of the camera and image acquisition of the workpiece.

In the image acquisition module, the left and right CCD industrial cameras 2 are fixed on the binocular camera fixture 11. There are guide rails on the fixture, and the left and right spacing and angle of the two CCD industrial binocular cameras 2 can be adjusted manually or by other tools. CCD industrial camera 2 is subject to binocular vision.

Specific implementation (as shown in Figure 3 and Figure 4):

As shown in Figure 5, the gray coordinate system (X, Y, Z) is the camera coordinate system, the two black two-dimensional coordinate systems are the pixel coordinate system (x, y) of the left and right cameras, and the pixel coordinates of the left camera The frame is located at (0,0,0) of the camera coordinate system, and the pixel coordinate system of the right camera is located at (Tx,0,0) of the camera coordinate system. (X, Y, Z) in the figure is a point in the camera coordinate system.

As shown in Figure 6, the workpiece coordinate system is established with the center of the bottom circle of the target workpiece as the base point.

As shown in Figure 7, the coordinate system of the manipulator is established with the origin of the center of the manipulator base.

Step 1: by adjusting the height of the industrial camera 2, the shooting angle and the distance between the two industrial cameras 2, the target workpiece 10 is located within the focal length range of the industrial camera 2;

The binocular camera fixture 11 moves up and down along the vertical rods on both sides of the camera bracket 3 to adjust the height of the industrial camera 2, and the two industrial cameras 2 move along the guide rail of the binocular camera fixture 11 to realize the distance between the two industrial cameras 2 Adjustment, the adjustment of the shooting angle of the industrial camera 2 is realized by adjusting the installation angle of the two industrial cameras 2 .

Step 2: According to the Zhang Zhengyou calibration method, perform double-target calibration on the left and right cameras to obtain the internal and external parameter matrix of the camera, and perform stereo calibration on the left and right cameras according to the parameter matrix to obtain a correction map.

Step 3: Select the hand-to-eye (Eye-To-Hand) hand-eye calibration mode for the four-axis manipulator 1 and the industrial camera 2 to perform hand-eye calibration, and obtain the pose transformation relationship between the camera coordinate system and the manipulator coordinate system;

Among them, the camera coordinate system is established by the left camera of the industrial camera as the origin, and the manipulator coordinate system is established by the base of the four-axis manipulator 1 as the origin.

Step 4: The target workpiece 10 to be grabbed is conveyed to the position below the binocular camera through the belt conveyor 9, and when the proximity sensor 8 receives the signal of the target workpiece 10, the belt conveyor motor drive device 4 stops the belt conveyor 9 jobs.

Step 5: The left and right cameras capture images of the target workpiece 10 passing the proximity sensor 8, and transmit the captured images of the target workpiece to the computer.

Step 6: Use the correction map obtained in step 1 to correct the target workpiece images collected by the left and right cameras, and then perform image segmentation, grayscale transformation, median filtering and edge detection preprocessing on the images in sequence.

Step 7: Use the NCC similarity measurement algorithm to perform stereo matching on the image preprocessed in Step 6; find matching points through the pyramid model of binocular stereo vision, and form a depth map from the point cloud image formed after the coordinates of all matching points are assembled. Calculate the average depth of the depth map to obtain the three-dimensional coordinates of the target workpiece 10 in the camera coordinate system;

The NCC similarity measurement algorithm can overcome the influence of uneven illumination in the image.

Step 8: Estimating the attitude of the target workpiece 10 based on Halcon to obtain the attitude information Q of the target workpiece 10;

First, use SolidWorks to draw a 3D image of the target workpiece 10. The pose of the target workpiece in the 3D image is an ideal pose; import the 3D image of the workpiece into Halcon and read it, and use the create_model_3d() operator in Halcon to generate a 3D shape Matching template, use the operator find_shape_model_3d() to match the preprocessed image obtained in step 5 with the 3D shape matching template, and obtain the attitude information Q of the target workpiece 10 relative to the manipulator coordinate system.

Step 9: Know the initial position and initial attitude of the end effector of the four-axis manipulator 1 (that is, the three-dimensional coordinates and attitude information of the end effector in the initial state are known), use the Cartesian position control mode for trajectory planning, and the four-axis manipulator 1 The end effector moves from the initial position to the vicinity of the target workpiece 10 according to the planned trajectory, and keeps the end effector always pointing to the target workpiece 10 during the movement, so that the grasping can be performed more efficiently;

The Cartesian position control mode performs trajectory planning according to the azimuth angle of the target workpiece 12 relative to the end effector and the three-dimensional coordinates of the target workpiece 12 in the camera coordinate system; the azimuth angle of the target workpiece 12 relative to the end effector is obtained by the following calculation: First, convert the three-dimensional coordinates of the target workpiece 10 in the camera coordinate system into the three-dimensional coordinates of the target workpiece 10 in the manipulator coordinate system through the pose conversion relationship in step 2, according to the three-dimensional coordinates of the target workpiece 10 and the end effector in the manipulator coordinate system The azimuth of the target workpiece 10 relative to the end effector is calculated.

Step 10: When the four-axis manipulator 1 approaches the target workpiece 10 during the grasping process, in order to prevent the manipulator from touching the workpiece to change its posture during the posture adjustment process, set a buffer distance d here, and start when the end of the manipulator reaches near d Adjust the attitude, and then reach the target point in a gradual manner to complete the grasping. The attitude of the end of the manipulator when grasping is Q. The attitude information Q of the target workpiece is transmitted to the control system of the four-axis manipulator 1 based on the TCP/IP protocol SOCKET communication, so that the four-axis manipulator completes the grasping of the target workpiece.

Adjust the attitude information of the end effector of the four-axis manipulator 1 according to the parameters m, n, and q. The calculation methods of the parameters m, n, and q are as follows:

As shown in Figure 8, in the manipulator coordinate system, set the counterclockwise rotation angle of the end effector of the four-axis manipulator 1 around the x-axis, y-axis, and z-axis as m, n, and q, respectively, and m, n, and q are the control The parameters of the end effector's up and down, left and right, and rotational movements;

Since the end effector always points to the object during the grasping process, the parameter m = α;

Among them, m is the rotation angle of the end effector around the x axis of the manipulator coordinate system based on the z axis, and α is the azimuth angle of the target workpiece relative to the end effector along the x axis;

The parameters n and q are calculated according to the attitude information Q of the target workpiece. The attitude information Q of the target workpiece includes the Euler angles θ _x , θ _y , and θ _z of the workpiece coordinate system rotating around the x, y, and z axes of the manipulator coordinate system. Thus, the rotation angle n of the end effector around the y-axis of the manipulator coordinate system is θ _y , and the rotation angle q around the z-axis of the manipulator coordinate system is θ _z :

That is, n = θ _y

q ₌ θz

The workpiece coordinate system is the coordinate system established with the target workpiece as the origin.

Step eleven: when the four-axis manipulator 1 grabs the target workpiece 10 , the posture information of the target workpiece 10 is the same as the posture information Q of the target workpiece 10 in step eight. The four-axis manipulator 1 first grabs the target workpiece 10 and moves up the set distance H along the Z-axis of the manipulator coordinate system, and then the four-axis manipulator 1 controls the end effector to move to the initial position, and then checks the attitude information Q and the end effector at the initial position. After the inverse kinematic solution is performed on the attitude information of the state, the end effector of the manipulator is adjusted back to the initial attitude of the end effector. At this time, the target workpiece 10 completes the transfer process from the initial attitude to the intermediate attitude.

Step 12: The four-axis manipulator 1 continues to grab the target workpiece 10 to the target stacking platform 6 , the target workpiece 10 is in the final posture, and the positioning-grabbing-placement process of the target workpiece 10 is completed.

Claims (7)

1. A device based on binocular vision pose recognition and grabbing, characterized in that it includes a workpiece delivery module, an image acquisition module and a target workpiece grabbing module; The workpiece transmission module includes a belt conveyor (9), a belt conveyor motor drive (4) and a proximity sensor (8), and a conveyor belt for transmitting a target workpiece (10) is installed on the belt conveyor (9). A proximity sensor (8) is installed on the frame of the belt conveyor (9) on one side of the conveyor belt; The image acquisition module includes an industrial camera (2), a camera bracket (3), a strip light source (5) and a binocular camera fixture (11), and the camera bracket is fixed on the frame of the belt conveyor (9) on both sides of the conveyor belt (3), a binocular camera fixture (11) is installed between the two sides of the camera bracket (3), two industrial cameras (2) are slidably installed on the guide rail of the binocular camera fixture (11), two industrial cameras (2 ) is used to photograph the target workpiece (10) passing through the proximity sensor (8), and the two industrial cameras (2) are respectively a left camera and a right camera; two symmetrically arranged strip light sources are installed on the bottom of both sides of the camera bracket (3) (5); the proximity sensor (8) is arranged near the camera bracket (3); The target workpiece grabbing module includes a four-axis manipulator (1) and a control cabinet (7). The control cabinet (7) is located on one side of the belt conveyor (9) and has the same height as the conveyor belt; the four-axis manipulator (1) passes through its base fixed on the control cabinet. 2. adopt the method for a kind of pose recognition and grasping based on binocular vision of device described in claim 1, it is characterized in that, comprise the following steps: Step 1: Adjust the height of the industrial camera (2), the shooting angle and the distance between the two industrial cameras (2), so that the target workpiece (10) is within the focal length range of the industrial camera (2); Step 2: According to the Zhang Zhengyou calibration method, perform double-target calibration on the left and right cameras to obtain the internal and external parameter matrix of the camera, and perform stereo calibration on the left and right cameras according to the parameter matrix to obtain the correction map; Step 3: Select the hand-to-eye (Eye-To-Hand) hand-eye calibration mode for the four-axis manipulator (1) and the industrial camera (2) to perform hand-eye calibration, and obtain the pose transformation relationship between the camera coordinate system and the manipulator coordinate system; Among them, the camera coordinate system is established by the left camera of the industrial camera as the origin, and the manipulator coordinate system is established by the four-axis manipulator (1) base as the origin; Step 4: The target workpiece (10) to be grasped is sent to the position below the binocular camera through the belt conveyor (9), and when the proximity sensor (8) receives the signal of the target workpiece (10), the belt conveyor motor The driving device (4) stops the work of the belt conveyor (9); Step 5: The left and right cameras capture the image of the target workpiece (10) passing through the proximity sensor (8), and transmit the captured image of the target workpiece to the computer; Step 6: Use the correction map obtained in step 1 to correct the target workpiece images collected by the left and right cameras, and then perform image segmentation, grayscale transformation, median filtering and edge detection preprocessing on the images in sequence; Step 7: Use the NCC similarity measurement algorithm to perform stereo matching on the image preprocessed in step 6, find matching points through the pyramid model of binocular stereo vision, and form a depth map from the point cloud image formed by gathering the coordinates of all matching points. Calculate the average depth of the depth map to obtain the three-dimensional coordinates of the target workpiece (10) in the camera coordinate system; Step 8: Perform attitude estimation on the target workpiece (10) based on Halcon to obtain the attitude information Q of the target workpiece (10); Step 9: Know the initial position and initial attitude of the end effector of the four-axis manipulator (1), and use the Cartesian position control mode for trajectory planning. The Cartesian position control mode is based on the azimuth angle of the target workpiece (12) relative to the end effector Trajectory planning is carried out with the three-dimensional coordinates of the target workpiece (12), the end effector of the four-axis manipulator (1) moves from the initial position to the vicinity of the target workpiece (10) according to the planned trajectory, and the end effector always points to the target workpiece during the movement process (10); Step 10: When the buffer distance between the end effector of the four-axis manipulator (1) and the target workpiece (10) is d, adjust the attitude information of the end effector of the four-axis manipulator (1), and then approach the target workpiece (10) in a gradual manner to complete the grasping Pick; Step 11: When the four-axis manipulator (1) grabs the target workpiece (10), the four-axis manipulator (1) first grabs the target workpiece (10) and moves up the set distance H along the Z axis of the manipulator coordinate system, and then The four-axis manipulator (1) controls the end effector to move to the initial position, and then solves the inverse kinematics for the attitude information Q and the attitude information of the end effector in the initial state, and then adjusts the end effector of the manipulator back to the initial attitude of the end effector. When the target workpiece (10) completes the transfer process from the initial posture to the intermediate posture; Step 12: The four-axis manipulator (1) continues to grab the target workpiece (10) to the target stacking platform (6), and the target workpiece (10) is in an end posture. 3. the method for a kind of pose recognition and grasping based on binocular vision according to claim 2, is characterized in that, described step eight is specifically: first utilize SolidWorks to carry out 3D image drawing to target workpiece (10); Import the 3D image of the workpiece into Halcon and read it, use the create_model_3d() operator in Halcon to generate a 3D shape matching template, use the operator find_shape_model_3d() to match the preprocessed image obtained in step 5 with the 3D shape matching template, Attitude information Q of the target workpiece (10) is obtained. 4. a kind of method based on binocular vision pose recognition and grasping according to claim 2, is characterized in that, in described step 9, target workpiece (12) with respect to the azimuth angle of end effector through following The calculation is obtained: firstly, the three-dimensional coordinates of the target workpiece (10) in the camera coordinate system are transformed into the three-dimensional coordinates of the target workpiece (10) in the manipulator coordinate system through the pose conversion relationship in step 2, and the target workpiece (10) and the end are executed The azimuth angle of the target workpiece (10) relative to the end effector is calculated based on the three-dimensional coordinates of the device in the manipulator coordinate system. 5. A method of pose recognition and grasping based on binocular vision according to claim 2, characterized in that, according to parameters m, n, q, adjust the position of the four-axis manipulator (1) end effector in step ten Attitude information, the calculation method of parameters m, n, q is as follows: In the manipulator coordinate system, set the counterclockwise rotation angles of the end effector of the four-axis manipulator (1) around the x-axis, y-axis, and z-axis as m, n, and q, and m, n, and q are the up and down directions of the control end effector, respectively. , left and right, rotation parameters; Since the end effector always points to the object during the grasping process, the parameter m = α; Among them, m is the rotation angle of the end effector around the x-axis of the manipulator coordinate system, and α is the azimuth angle of the target workpiece relative to the end effector along the x-axis; The parameters n and q are calculated according to the attitude information Q of the target workpiece, which includes the Euler angles θ _x , θ _y , and θ _z of the target workpiece rotating around the x, y, and z axes of the manipulator coordinate system, so that The rotation angle n of the end effector around the y-axis of the manipulator coordinate system is θ _y , and the rotation angle q around the z-axis of the manipulator coordinate system is θ _z : That is, n = θ _y q ₌ θz The workpiece coordinate system is the coordinate system established with the target workpiece as the origin. 6. A method of pose recognition and grabbing based on binocular vision according to claim 2, characterized in that the pose information when the end effector of the four-axis manipulator (1) grabs the target workpiece (10) It is the same as the attitude information Q of the target workpiece (10) in step eight. 7. A method of pose recognition and grasping based on binocular vision according to claim 2, characterized in that the pose information Q of the target workpiece is transmitted to the four-axis based on TCP/IP protocol SOCKET communication A control system for a manipulator (1).