A Vision-based Autonomous and Precise Positioning Method and Device for PCB Sheets

Abstract

The present invention discloses a method and device for autonomous and precise positioning of PCB boards based on vision. Aiming at the characteristics of unprocessed PCB boards, such as large area, single feature, and large self-shape and size, traditional vision processing is difficult to effectively solve the problem that precise positioning cannot be achieved due to the large deviation of the camera from the robot's rotation axis, which amplifies the error during robot loading and unloading. This method uses a high-precision calibration board as a positioning marker target to solve the positioning and recognition problems of large positioning objects without obvious features, effectively overcoming the problem of precise positioning that is difficult for existing vision detection technologies when the PCB board has a large area and no obvious features; constructing an error compensation model for correction, multiple compensation corrections can be performed to make the model accuracy reach a practical purpose; the positioning accuracy is further improved by using an iterative method.

Description

Technical Field

[0001] The present invention belongs to the technical field of automatic loading and unloading and self-positioning, and particularly relates to a method and device for automatic and precise positioning of PCB boards based on vision. Background Art

[0002] In a PCB automatic production line, for the loading and unloading of PCB boards, in order to ensure the consistency of the positions of the machine tool coordinate system and the tool coordinate system, complex clamping adjustments are required, with complex steps, long operation time, and low efficiency. In addition, there are various types of PCB boards to be processed, with large board areas, heavy weights, no obvious features, and high precision requirements for the positioning of the processing table. At present, the PCB board loading and unloading positioning technology in the automatic production line mainly focuses on static positioning and grasping. The poses of the robotic arm and the camera itself and the positions of the grasped target objects need to maintain a certain relative relationship. The visual recognition and positioning types of the end effector are single and have poor adaptability, and cannot adapt to the dynamic and precise positioning in variable and complex application scenarios, and cannot meet the current production requirements of high precision, high stability, and high adaptability. During the positioning process of the PCB board, the poses of the robot and the camera itself are constantly changing. For example, when the robotic arm picks up and delivers objects at fixed points during loading and unloading, the uncertainty of its own repeated positioning affects the grasping accuracy, and the offset between the material rack and the new batch of materials after material replacement also affects the grasping and positioning. To achieve the universality and dynamics of the grasped object, the PCB board, and improve the grasping accuracy, robot vision technology can be used to obtain information about the workpiece and its surrounding environment, and an external positioning marker target is used to endow features to help identify the target workpiece to be operated and improve the positioning and recognition accuracy, guiding the industrial robot to complete operations such as grasping and placing the workpiece.

[0003] In traditional PCB board positioning methods, there are mainly the following three: First, technicians use manual or robotic arm teaching positioning to position the workpiece to be processed, which is easily affected by human subjective factors. The positioning of the target workpiece is also affected by collisions and frictions during handling, with poor positioning effects and low accuracy. Second, the robotic arm teaching method is used to perform repeated operations according to the manually specified path trajectory, which requires the fixed unity of the grasped position and the placed position of the target, has certain requirements for the environment, and lacks flexibility. Third, visual positioning technology is used to extract the feature points of the workpiece, and the position of the workpiece is determined by means of template matching, feature data verification, etc., and then precise processing is carried out. However, this often targets the features of a single object, and the lack of obvious prominent features of the PCB board results in low recognition accuracy and inability to identify using this method. Summary of the Invention

[0004] Aiming at the deficiencies of the existing PCB blank positioning detection, the purpose of the embodiments of the present application is to provide a vision-based autonomous and precise positioning method and device for PCB blanks. The vision positioning technology is used to determine the position of the workpiece. The automatic positioning of the PCB can solve various subjective problems in manual positioning, quickly provide accurate size and position information, and then perform precise processing, which has significant advantages in terms of accuracy and efficiency.

[0005] According to the first aspect of the embodiments of the present application, a vision-based autonomous and precise positioning method for PCB blanks is provided, including:

[0006] Step (1): After installing a camera support module at the end of the robotic arm and fixedly installing a camera at the end of the L-shaped camera fixing part in the camera support module, and the robotic arm clamps the PCB blank, obtain the pre-set operating path of the robotic arm;

[0007] Step (2): Obtain the distorted calibration plate image taken under the pre-set camera parameters, and solve the distortion coefficient according to the distorted calibration plate image;

[0008] Step (3): Obtain the standard position image where the positioning mark target is located at the center of the camera's field of view, perform distortion correction on the standard position image using the distortion coefficient and then perform pre-processing to obtain the positioning mark target image area, extract the feature points of the positioning mark target, calculate the coordinate value and vector value of the center pixel point of the mark target in the image coordinate system, and set the position of the end of the robotic arm when the positioning mark target is at the center of the camera's field of view as the origin in the processing table coordinate system;

[0009] Step (4): Obtain the standard offset image set collected under the pre-set camera translation and rotation parameters, solve the image coordinate value of the center pixel point of the mark target, record the corresponding relationship between the pre-set standard offset and the image coordinate offset value of the mark target center point, and establish an error compensation model;

[0010] Step (5): Obtain and solve the image coordinate value of the target feature point of the image to be solved, compare it with the difference in the center pixel of the positioning mark target in the standard offset image set in step (4), perform interpolation between the two closest translation values and rotation values to obtain the actual offset, and drive the fixture to correct the offset according to the actual offset;

[0011] Step (6): After the fixture corrects the offset, repeat step (5) iteratively until the difference between the recognized positioning mark target image and the standard position image is less than the pre-set target accuracy range, and control the robotic arm to return to the standard position according to the offset value and drive the fixture to complete the loading and unloading actions.

[0012] Further, in the step (1), the process of "installing a camera support module at the end of the robotic arm and fixedly installing a camera at the end of the camera support, and the robotic arm clamping the PCB board" is as follows:

[0013] A square connecting piece is arranged below the flange connecting shaft at the end of the robotic arm and fixedly connected to the flange connecting shaft. One side of the L-shaped camera fixing piece is installed on the flange connecting shaft, one side of the rib support piece is fixedly connected to the flange connecting shaft through the square connecting piece, the camera is fixedly installed on the other side of the L-shaped camera fixing piece, and a fixture is installed directly below the flange connecting shaft to clamp the PCB board;

[0014] Wherein the axis of the flange connecting shaft coincides with the central axis of the robotic arm, the central axis of the PCB board deviates from the machine central axis within 5 mm, and the camera support deviates from the machine central axis by more than 280 mm.

[0015] Further, the step (2) includes:

[0016] Step 2.1: At a pre-set camera shooting height H, control the camera to shoot an image of the distortion calibration board, wherein the size of the distortion calibration board is larger than the shooting field of view of the camera;

[0017] Step 2.2: Extract the sub-pixel level feature points in the image of the distortion calibration board, and use the image coordinate values and actual coordinate values of the feature points to calculate the distortion coefficients by combining two types of radial and tangential distortions:

[0018]

[0019] In the formula, k1, k2, k3, p1, and p2 are five distortion coefficients, x and y are the coordinates of the feature points in the image of the distortion calibration board, x correted , y correted are the pixel coordinates of the corrected points.

[0020] Further, the step (3) includes:

[0021] Step 3.1: Fix the positioning mark target on the processing table, make the vision center position of the positioning mark target at the fixed shooting point, and control the camera to shoot an image of the standard position at the fixed shooting point, wherein the positioning mark target is a small calibration board with an area less than 1 / 9 of the shooting field of view within the shooting field of view;

[0022] Step 3.2: Distortion-correct the standard position image using the distortion coefficients, and sequentially perform histogram normalization, median filtering, brightness correction, and mean binarization on the corrected image, and obtain the positioning mark target image area using area and perimeter features;

[0023] Step 3.3: Solve the coordinates of all feature points in the image coordinate system and use the mean value of the coordinates of all feature points as the coordinates (x c , y c ) of the central pixel point of the positioning marker target:

[0024]

[0025] where (x pab , y pab ) are the sub-pixel feature point coordinates in the image coordinate system, a and b represent the row and column numbers of the feature points in the positioning marker target, 1 ≤ a ≤ n, 1 ≤ b ≤ n;

[0026] Step 3.4: Fit to obtain the vectors x1,..., x n of each row of feature points in the image coordinate system, and solve the vector mean as the vector value of the central pixel point of the positioning marker target;

[0027] Step 3.5: Set the position of the end of the robotic arm when the positioning marker target is at the center of the camera's field of view as the origin in the processing table coordinate system.

[0028] Further, the step (4) includes:

[0029] Step 4.1: With the origin in the processing table coordinate system as the center, control the actuator to perform a ±c mm translation with the xy direction in the tool coordinate system of the robotic arm as the reference direction, and control the camera to take the first picture set at the coordinates (-c, -c), (0, -c), (c, -c), (-c, 0), (0, 0), (c, 0), (-c, c), (0, c), (c, c) respectively. Let the number of shooting times be n 2 times, and the short side distance of the shooting field of view is l mm, where n ≥ 3, l / 2n ≤ c ≤ l / n mm;

[0030] Step 4.2: With the origin in the processing table coordinate system as the center, control the actuator to perform a ±d° rotation with the z axis in the tool coordinate system of the robotic arm as the rotation axis, and control the camera to take the second picture set at the rotation angles of -nd°... -3d°... 0°... 3d°... n d° respectively. Let the number of shooting times be 2n + 1, where n ≥ 1, 45 / (2n + 1)° ≤ d ≤ 90 / (2n + 1)°;

[0031] Step 4.3: Obtain the central point pixel coordinates and coordinate systems of the positioning marker target in the first picture set and the second picture set, and according to the corresponding translation or rotation coordinate actual variables, obtain the translation vector and rotation vector of the processing table coordinate system offset and the pixel coordinate system offset;

[0032] Step 4.4: Record the actual offset value of the robotic arm and the offset value calculated by vision, collect the error data at each position, and establish an error compensation model;

[0033] Further, the step (5) includes:

[0034] Step 5.1: Run the end of the robotic arm to the fixed shooting point to take a picture, and calculate the pixel coordinate system of the center point of the marking target;

[0035] Step 5.2: Taking the pixel coordinate system of the center point of the marking target in the standard position as the reference value, obtain the pixel offset value of the center point of the marking target to be recognized relative to the standard value;

[0036] Step 5.3: Take the two closest sets of pixel offset values, interpolate and solve the translation amount and rotation amount between the two sets of data, and convert the pixel offset value into an actual offset amount;

[0037] Step 5.4: According to the error compensation model, obtain the error value in the error compensation model corresponding to the pixel coordinates of the center point of the marking target, and correct the actual offset amount;

[0038] Step 5.5: Drive the fixture to correct the offset according to the corrected offset amount.

[0039] Further, the step (6) includes:

[0040] Step 6.1: Repeat step (5) to obtain the corrected offset amount;

[0041] Step 6.2: Compare whether the corrected offset amount is less than the required target accuracy range. If so, send a True instruction to the robot, make the end of the robotic arm return to the standard position in step 3.1 according to the corrected offset amount, and drive the fixture to complete the loading and unloading action. Otherwise, send a False to the actuator, perform step (5) again, identify the current corrected offset value, and make the upper computer send the corrected offset amount to the actuator for iterative adjustment.

[0042] According to the second aspect of the embodiments of the present application, a vision-based PCB board material autonomous precise positioning device is provided, including:

[0043] An acquisition module, configured to install a camera support module at the end of the robotic arm and fixedly install a camera at the end of the L-shaped camera fixing part in the camera support module, and obtain a pre-set robotic arm operation path after the robotic arm grips the PCB board;

[0044] A solution module, configured to obtain a distorted calibration plate image taken under pre-set camera parameters, and solve the distortion coefficient according to the distorted calibration plate image;

[0045] A setting module, configured to obtain a standard position image of a positioning marker target located at the exact center of the camera's field of view, perform preprocessing on the standard position image after distortion correction using the distortion coefficient to obtain a positioning marker target image region, extract feature points of the positioning marker target, calculate the coordinate value and vector value of the center pixel point of the marker target in the image coordinate system, and set the position of the end of the robotic arm when the positioning marker target is located at the exact center of the camera's field of view as the origin in the processing table coordinate system;

[0046] A modeling module, configured to obtain a set of standard offset images collected under preset camera translation and rotation parameters, solve the image coordinate value of the center pixel point of the marker target, record the correspondence between the preset standard offset and the image coordinate offset value of the center point of the marker target, and establish an error compensation model;

[0047] A correction module, configured to obtain and solve the image coordinate value of the feature point of the image target to be solved, compare it with the difference in the center pixel of the positioning marker target in the standard offset image set in step (4), interpolate between the two closest translation values and rotation values to obtain the actual offset, and drive the fixture to correct the offset according to the actual offset;

[0048] A control module, configured to, after the fixture corrects the offset, repeatedly execute the steps in the correction module iteratively until the difference between the recognized positioning marker target image and the standard position image is less than a preset target accuracy range, control the robotic arm to return to the standard position according to the offset value, and drive the fixture to complete the loading and unloading actions.

[0049] According to the third aspect of the embodiments of the present application, there is provided an electronic device, including:

[0050] One or more processors;

[0051] A memory, configured to store one or more programs;

[0052] When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in the first aspect.

[0053] According to the third aspect of the embodiments of the present application, there is provided a computer-readable storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method as described in the first aspect are implemented.

[0054] The technical solutions provided by the embodiments of the present application may include the following beneficial effects:

[0055] As can be seen from the above embodiments, the present application combines machine vision to enable the robotic arm to achieve autonomous and automated loading and unloading, effectively overcoming the problems of friction and collision of the material board caused by manual positioning of traditional PCB material boards, and the problem that existing vision detection is difficult to perform dynamic and precise positioning in complex environments; the method of using a high-precision calibration board as a positioning marker target solves the positioning and recognition problems of large positioning objects without obvious features. By taking advantage of the high precision of the calibration board, obvious and regular features of the marker target are recognized, which not only improves the accuracy of positioning and recognition to a high degree, but also greatly weakens the influence of complex environments on the recognition stability. It is not limited to a single object of the PCB material board, improving the generalization rate of the recognition object. In addition, the method based on iterative adjustment can achieve autonomous setting of the target accuracy, further improving the positioning accuracy.

[0056] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The accompanying drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application.

[0058] Figure 1 is a flowchart of a method for autonomous and precise positioning of PCB board materials based on vision shown according to an exemplary embodiment.

[0059] Figure 2 is a diagram of a camera bracket shown according to an exemplary embodiment.

[0060] Figure 3 is a schematic configuration diagram of a PCB material board shown according to an exemplary embodiment.

[0061] Figure 4 is a schematic diagram of the positioning recognition and correction position of a PCB material board shown according to an exemplary embodiment.

[0062] Figure 5 is a diagram of a distortion and distortion correction shown according to an exemplary embodiment.

[0063] Figure 6 is a diagram of a standard translation offset shown according to an exemplary embodiment.

[0064] Figure 7 is a diagram of a standard rotation offset shown according to an exemplary embodiment.

[0065] Figure 8 is a block diagram of a device for autonomous and precise positioning of PCB board materials based on vision shown according to an exemplary embodiment.

[0066] Figure 9It is a schematic diagram of an electronic device shown according to an exemplary embodiment.

[0067] In the figure, 1 is the camera support module; 11 is the camera; 12 is the rib support; 13 is the square connecting piece; 14 is the flange connecting shaft; 15 is the L-shaped camera fixing piece; 16 is the fixture; 17 is the PCB board; 18 is the processing table; 19 is the positioning pin; 20 is the target pin hole; 21 is the positioning marker target; 22 is the robotic arm. Detailed implementation manners

[0068] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present application.

[0069] The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The singular forms "a", "the" and "said" used in the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0070] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0071] Figure 1 It is a flowchart of a vision-based autonomous precise positioning method for PCB boards according to an exemplary embodiment, as Figure 1 shown, and the method may include the following steps:

[0072] Step (1): After installing the camera support module 1 at the end of the robotic arm 22 and fixedly installing the camera 11 at the end of the L-shaped camera fixing piece 15 in the camera support module 1, and the robot gripping the PCB board 17, obtain the pre-set operating path of the robotic arm 22;

[0073] Step (2): Obtain the distorted calibration board image captured under the pre-set camera 11 parameters, and solve the distortion coefficient according to the distorted calibration board image;

[0074] Step (3): Obtain the standard position image of the positioning marker target 21 at the exact center of the camera 11's field of view. After performing distortion correction on the standard position image using the distortion coefficient, preprocess it to obtain the image area of the positioning marker target 21, extract the feature points of the positioning marker target 21, calculate the coordinate value and vector value of the center pixel point of the marker target in the image coordinate system, and set the position of the end of the robotic arm 22 when the positioning marker target 21 is at the exact center of the camera 11's field of view as the origin in the coordinate system of the processing table 18;

[0075] Step (4): Obtain the standard offset image set collected under the pre-set camera 11 translation and rotation parameters, solve the image coordinate value of the center pixel point of the marker target, record the corresponding relationship between the pre-set standard offset and the image coordinate offset value of the marker target center point, and establish an error compensation model;

[0076] Step (5): Obtain and solve the image coordinate value of the target feature point of the image to be solved, compare it with the difference in the center pixel of the positioning marker target 21 in the standard offset image set in step (4), interpolate between the two closest translation values and rotation values to obtain the actual offset, and drive the fixture 16 to correct the offset according to the actual offset;

[0077] Step (6): After the fixture 16 corrects the offset, repeat step (5) iteratively until the difference between the recognized image of the positioning marker target 21 and the standard position image is less than the pre-set target accuracy range, control the robotic arm 22 to return to the standard position according to the offset value and drive the fixture 16 to complete the loading and unloading operation.

[0078] As can be seen from the above embodiments, this application combines machine vision to enable the robotic arm 22 to achieve autonomous and automated loading and unloading, effectively overcoming the problems of friction and collision of the material board during manual positioning of the traditional PCB material board 17, and the problem that existing vision detection is difficult to perform dynamic and precise positioning in complex environments; the method using a high-precision calibration board as the positioning marker target 21 solves the positioning and recognition problem of large positioning objects without obvious features. Utilizing the high-precision advantage of the calibration board to identify the obvious and regular features of the marker target not only improves the positioning and recognition accuracy to a high degree but also greatly weakens the influence of complex environments on the recognition stability. It is not limited to a single object such as the PCB material board 17, improving the generalization rate of the recognized objects. In addition, the method based on iterative adjustment can achieve autonomous setting of the target accuracy, further improving the positioning accuracy.

[0079] Refer to Figure 1, the implementation process of the present invention is as follows: First, determine the shooting parameters to shoot a calibration target plate larger than the field of view, and perform precise distortion correction using the standard parameters of the calibration target plate. Secondly, select a small high-precision feature calibration target plate as the positioning marker target 21 according to the field of view. When the high-precision feature calibration target plate is placed at the center of the field of view, it is the standard position. Then move the PCB board 17, the fixture 16, and the camera 11, and collect the standard offset map according to the change rate of the fixed translation variable a mm and the rotation variable °, and obtain the transformation relationship between the pixel offset value and the actual offset variable. Then take the original image of the positioning marker target 21 to be recognized, identify the pixel coordinates of the feature points, solve the actual offset amount according to the standard offset value, and make the PCB board and the camera 11 adjust the offset amount. Finally, take a new image of the positioning marker target 21 after adjustment, and judge whether the current offset value is greater than the precision range b mm. If so, perform a second adjustment; otherwise, let the PCB board run to the processing table 18 according to the specified plan. In an embodiment, a = 40mm; a = 5°; b = 0.5mm. Those skilled in the art can set them according to the actual situation and will not be elaborated here.

[0080] In the specific implementation of step (1), the process of "installing the camera support module 1 at the end of the robotic arm 22 and fixedly installing the camera 11 at the end of the camera 11 support, and the robotic arm 22 clamping the PCB board 17" is as follows:

[0081] A square connecting piece 13 is arranged below the flange connecting shaft 14 at the end of the robotic arm 22, fixedly connected to the flange connecting shaft 14. One side of the L-shaped camera fixing piece 15 is installed on the flange connecting shaft 14. One side of the rib support piece is fixedly connected to the flange connecting shaft 14 through the square connecting piece 13. The camera 11 is fixedly installed on the other side of the L-shaped camera fixing piece 15. A fixture 16 is installed directly below the flange connecting shaft 14 to clamp the PCB board 17;

[0082] The axis of the flange connecting shaft 14 coincides with the central axis of the robotic arm 22. The central axis of the PCB board 17 deviates from the machine central axis within 5 mm, and the camera 11 support deviates from the machine central axis by more than 280 mm.

[0083] Refer to Figure 2 , which is a schematic diagram of the camera 11 support of the present invention, specifically including four parts: a rib support rod, a square connecting piece 13, a flange connecting shaft 14, and an L-shaped camera fixing piece 15. Except for the flange connecting shaft 14, the other parts are made of stainless steel alloy with a thickness of 3-4 mm. The rib support rod adopts a triangular structure and an optimized hollow structure, which not only enhances the stability after the camera 11 is installed but also reduces the structure weight, realizing a lightweight structure to reduce the load of the robotic arm 22.

[0084] The L-shaped camera fixing part 15 has a slot in the middle, which is convenient for installation and positioning with the rib support rod, and symmetric holes are provided for fixing the cable of the camera 11. After positioning and installation, the L-shaped camera fixing part 15 and the rib part are spot-welded and fixed around the tenon groove; a square connecting part 13 is welded below the flange connecting shaft 14, and the rib support rod and the flange connecting shaft 14 can be fixedly connected through the square connecting part 13 to avoid the offset between the rib support rod and the flange connecting shaft 14. A clamp 16 is installed directly below the flange connecting shaft 14 to clamp the PCB board 17. The axis of the flange connecting shaft 14 coincides with the central axis of the end of the robotic arm 22, the center of the PCB board 17 is close to coinciding with the central axis of the end of the robotic arm 22, and the camera 11 bracket deviates from the central axis of the end of the robotic arm 22. Manually configure the key points in the running path of the robot to avoid the collision and interference between the board and the camera 11 during operation with the workbench machine tool.

[0085] Using the designed camera 11 bracket structure, the jitter of the camera 11 is reduced, the stability of the image collected by the loading and unloading camera 11 is increased, and at the same time, a lightweight structure is realized, overcoming the problem of blurred image features caused by the jitter of the camera 11 during image acquisition;

[0086] Refer to Figure 3 , which is the overall configuration schematic diagram of the PCB board 17 of the present invention, specifically including a camera bracket module 1, a camera 11, a rib support member 12, a square connecting member 13, a flange connecting shaft 14, an L-shaped camera fixing part 15, a clamp 16, a PCB board 17, a processing table 18, a positioning pin 19, a target pin hole 20, a positioning mark target 21, and a robotic arm 22; among them, the positioning pin 19 needs to be accurately placed in the pin hole, and the positioning accuracy is guaranteed within ±0.1 mm, and each component will not interfere during the operation process, and the position of the positioning mark target 21 is fixed at a position that does not affect the processing of the PCB board 17.

[0087] In the specific implementation of step (2), step (2) includes:

[0088] Step 2.1: At the preset shooting height H of the camera 11, control the camera 11 to shoot the image of the distortion calibration board, wherein the size of the distortion calibration board is larger than the shooting field of view range of the camera 11;

[0089] Specifically, refer to Figure 4, which is a schematic diagram of the positioning, identification, and correction of the PCB blank 17 of the present invention. The relative position between the camera 11 and the PCB blank 17 remains fixed. Determine the shooting height H in the figure to ensure that the standard position of the positioning mark target 21 is at the center of the field of view, and ensure that the field of view can capture the mark target within the dynamic change range of the camera 11 and the PCB blank 17. In this application, the method of using a high-precision calibration plate as the positioning mark target 21 solves the positioning and identification problems of large positioning objects without obvious features. By utilizing the high-precision advantage of the calibration plate, the obvious and regular features of the mark target are identified, which not only improves the positioning and identification accuracy to a high degree but also greatly weakens the influence of the complex environment on the recognition stability. It is not limited to the single object of the PCB blank 17, and the generalization rate of the recognition object is improved.

[0090] Refer to Figure 5 , which are the distortion map and the distortion correction map of the present invention. First, select a calibration plate larger than the field of view according to the shooting field of view of the camera 11 as the positioning mark target 21, make the mark target fill the entire image, and control the camera 11 to capture the distorted calibration plate image in this state.

[0091] Step 2.2: Extract the sub-pixel level feature points in the distorted calibration plate image. Using the image coordinate values and actual coordinate values of the feature points, combine the two types of radial and tangential distortions to obtain the distortion coefficients:

[0092]

[0093] In the formula, k1, k2, k3, p1, and p2 are five distortion coefficients, x and y are the pixel coordinates of the feature points in the distorted calibration plate image, x correted , y correted are the pixel coordinates of the feature points after distortion correction.

[0094] Specifically, after preprocessing the distorted calibration plate image, extract the sub-pixel level feature points of the distorted calibration plate. Use the Huber weight function to iteratively fit the feature points into a straight line, and combine the two types of radial and tangential distortions to obtain the distortion coefficients.

[0095] Substitute Figure 5 the feature point data in into the above formula, and the distortion coefficients can be obtained by solving:

[0096] dist = [-0.07618824 0.09408335 -0.00011556 0.00051624 -0.05627428]

[0097] In the specific implementation of step (3), the said step (3) includes:

[0098] Step 3.1: Fix the positioning marker target 21 on the processing table 18, making the marker target at the center position of the field of view under the fixed shooting point. Control the camera 11 to take an image of the standard position under the fixed shooting point, where the positioning marker target 21 is a small calibration plate with an area less than 1 / 9 of the field of view within the shooting field of view.

[0099] Specifically, fix a calibration plate with an area of 40mm 2 on the non-working area of the tabletop of the processing table 18 according to the actual situation. According to the hardware parameters of the 5-million-pixel industrial camera 11 and the 8mm focal length lens of the application device, adjust the vertical shooting distance of the camera 11 to 20cm, making the calibration plate within the depth of field of the camera 11 and physically close to the center of the camera 11's field of view. Finally, record the parameters of the robotic arm 22 at this position, and use the position of the robotic arm 22 under this parameter as the fixed shooting point.

[0100] Step 3.2: Use the distortion coefficient to correct the distortion of the standard position image. Successively perform histogram normalization, median filtering, brightness correction, and mean binaryzation on the corrected image, and obtain the image area of the positioning marker target 21 using area and perimeter features.

[0101] In an embodiment, histogram normalization with a gray range of 0 - 255, median filtering with a kernel size of 5×5, brightness correction with a gain of 6, and mean binaryzation with a kernel size of 3×3 can be successively performed on the corrected image. BLOB analysis, template matching, or watershed algorithm can be performed using area and perimeter features to obtain the image area of the positioning marker target 21. It should be noted that the parameters in the above histogram normalization, median filtering, brightness correction, and mean binaryzation processes can be obtained by those skilled in the art through experimental simulation, and the foregoing parameters are not the only parameter settings for this step.

[0102] Step 3.3: Solve the coordinates of all feature points in the image coordinate system and use the mean of the coordinates of all feature points as the coordinates (x c , y c ) of the central pixel point of the positioning marker target 21:

[0103]

[0104] where (x pab , y pab ) are the sub-pixel feature point coordinates in the image coordinate system, a and b represent the row and column numbers of the feature points in the positioning marker target 21, 1 ≤ a ≤ n, 1 ≤ b ≤ n;

[0105] Step 3.4: Fit to obtain the vectors x1,..., x n of each row of feature points in the image coordinate system, and solve the vector mean The vector value of the central pixel point for positioning the target 21 of the positioning marker;

[0106] Specifically, the least squares method, the Gauss-Newton method, or the Hough transform can be used for fitting. The feature points are grouped by row, and the pixel coordinate values of the feature points in the x-axis direction of the image coordinate system of each group are fitted. Finally, the vectors of each row of feature points in the image coordinate system are obtained.

[0107] Step 3.5: Take the coordinates and vector values of the central pixel point as standard values, and set the position of the end of the robotic arm 22 when the positioning marker target 21 is at the center of the field of view of the camera 11 as the origin in the coordinate system of the processing table 18.

[0108] In this application, all coordinate systems are based on the right-hand rule. The origin of the coordinate system of the processing table 18 is located at the center of the position of the punctuation board. The positive direction of the X-axis points to the center of the position where the material board is placed. The positive direction of the Y-axis points to the direction of the robot. The positive direction of the Z-axis points above the robot. The origin of the tool coordinate system is located at the center of the camera 11 in the end effector of the robotic arm 22. The positive direction of the X-axis points to the center of the flange connecting shaft 14 at the end of the robotic arm 22. The positive direction of the Y-axis points to the direction of the robot base. The positive direction of the Z-axis points above the robot. The three coordinate axes are respectively parallel to the three coordinate axes of the coordinate system of the processing table 18. The coordinate system of the processing table 18 remains unchanged after being set. The relationship between the tool coordinate system and the coordinate system of the processing table 18 will change with the position of the robotic arm 22 and the rotation of the joints. The image coordinate system is a two-dimensional coordinate. The upper left vertex of the acquired image is used as the origin of the coordinate system. The positive direction of the X-axis points to the right side in the horizontal direction of the image. The positive direction of the Y-axis points to the lower side in the vertical direction of the image.

[0109] In the specific implementation of step (4), step (4) includes:

[0110] Step 4.1: Centered at the origin in the coordinate system of the processing table 18, control the actuator to perform a ±c mm translation with the xy direction in the tool coordinate system of the robotic arm 22 as the reference direction, and control the camera 11 to take the first set of pictures at the coordinates (-c, -c), (0, -c), (c, -c), (-c, 0), (0, 0), (c, 0), (-c, c), (0, c), (c, c) respectively. The number of shooting times is n 2 times, the short side distance of the shooting field of view is l mm, n≥3, l / 2n≤c≤l / n mm;

[0111] In one embodiment, according to the short side distance of the field of view being 150 mm, a single translation amount of 40 mm is selected, and translations of ±40 mm are performed with the xy directions in the tool coordinate system of the robotic arm 22 as the reference directions. A picture is taken each time after translation at coordinates (-40, -40), (0, -40), (40, -40), (-40, 0), (0, 0), (40, 0), (-40, 40), (0, 40), and (40, 40). A total of 9 pictures are taken in 9 acquisitions to form the first picture set.

[0112] Step 4.2: With the origin in the coordinate system of the processing table 18 as the center, control the actuator to rotate by ±d° with the z-axis in the tool coordinate system of the robotic arm 22 as the rotation axis. Control the camera 11 to take pictures at rotation angles of -nd°…-3d°…0°…3d°…n d° respectively to obtain the second picture set. The number of acquisitions is 2n + 1, where n ≥ 1 and 45 / (2n + 1)° ≤ d ≤ 90 / (2n + 1)°.

[0113] In one embodiment, the standard position angle is selected as 0°, and image acquisition is performed with a single rotation angle of 5° around the z-axis in the tool coordinate system of the robotic arm 22. Pictures are taken from -20° to 20°, one picture every 5°, and a total of 9 pictures are taken in 9 acquisitions.

[0114] Step 4.3: Obtain the center point pixel coordinates and coordinate systems of the positioning marker target 21 in the first and second picture sets, and according to the corresponding actual variables of the translation or rotation coordinates, obtain the translation vector and rotation vector of the coordinate system offset of the processing table 18 and the pixel coordinate system offset.

[0115] Step 4.4: Record the actual offset value of the robot and the visually calculated offset value, calculate the error data at each position, and establish an error compensation model.

[0116] In one embodiment, take the offset values in the x and y offset directions as the two independent variables of the function, and use the least squares method to configure and fit a cubic surface equation of a binary function. The fitting formula is as follows:

[0117] E rror =K·W

[0118] =(k0 k1 k2 k3 k4 k5 k6 k7 k8 k9)·(1 x y xy x 2 y 2 xy 2 yx 2 x 3 y 3 ) T

[0119] =k0+k1x+k2y+k3xy+k4x2 +k5y 2 +k6xy 2 +k7yx 2 +k8x 3 +k9y 3 (3)

[0120] Among them, K is a set of coefficients, W is a set of independent variable functions, k0, …, k9 are the coefficients of the fitting surface equation, and x and y are the coordinate system variables of the processing table 18.

[0121] In specific implementation, Gaussian process fitting and kernel regression fitting methods can also be adopted to establish an error compensation model.

[0122] In the specific implementation of step (5), step (5) includes:

[0123] Step 5.1: Run the end of the robotic arm 22 to the position of the fixed shooting point to take a picture, and calculate the pixel coordinate system of the center point of the marking target.

[0124] Specifically, let the pose parameters of the robotic arm 22 run to the pose parameters recorded at the fixed shooting point. At this position, the robotic arm 22 sends a shooting signal to the upper computer to command the camera 11 to collect the calibration plate image at this time, and perform feature point extraction and other processing on this image to calculate the pixel coordinate value (1500, 1300) in the center point image coordinate system.

[0125] Step 5.2: Taking the pixel coordinate system of the center point of the marking target at the standard position as the reference value, obtain the pixel offset value of the center point of the marking target to be recognized relative to the reference value.

[0126] Specifically, taking the pixel coordinates (1239.786, 1027.382) of the center point of the marking target at the standard position as the reference value, obtain the difference between the pixel coordinate value of the center point of the calibration plate in the image taken in step 5.1 and the reference value. The difference is the pixel offset value (260.214, 272.618).

[0127] Step 5.3: Select the two closest sets of pixel offset values, interpolate and solve for the translation amount and rotation amount between the two sets of data, and convert the pixel offset value into an actual offset amount.

[0128] Specifically, assuming that the obtained pixel offset value is (260.214, 272.618), two sets of fixed offset coordinates in the x direction that are close on the corresponding x coordinate are 0 and 540.201, corresponding to offsets of 0 and 40 mm, and two sets of fixed offset coordinates in the y direction that are close on the y coordinate are 0 and 542.882, corresponding to offsets of 0 and 40 mm. Through interpolation calculation, the actual offset in the x direction is 19.27 mm, and the actual offset in the y direction is 20.09 mm. Similarly, assuming that the obtained x-direction vector fitting is (0.9, 0.46), the two closest vectors are (1, 0) and (0.99, 0.14), and the interpolation gives a rotation amount of 27.07°.

[0129] Step 5.4: According to the error compensation model, obtain the error value in the error compensation model corresponding to the pixel coordinates of the center point of the marking target, and correct the actual offset.

[0130] Specifically, according to the fitted error compensation model, the error values in the xy directions corresponding to the pixel coordinate values (1500, 1300) of the center point in the image coordinate system are (0.011, 0.185). The error is corrected. After correction, the total required error in the x direction is 19.27 + 0.011 = 19.281 mm, the total required error in the y direction is 20.09 + 0.185 = 20.275 mm, and the total angle error is 27.02 + 0.054 = 27.074°.

[0131] Step 5.5: According to the corrected offset, drive the fixture 16 to correct the offset.

[0132] Specifically, according to the offset corrected by the error compensation model, the robot is made to move relatively with the tool coordinate system as the reference, correct 19.281 mm in the x direction, correct 20.275 mm in the y direction, and correct 27.074° by rotating around the z axis.

[0133] In the specific implementation of step (6), step (6) includes:

[0134] Step 6.1: Repeat step (5) to obtain the corrected offset.

[0135] Specifically, return to step (5.1) to step (5.4) again, capture the calibration plate image at the measurement point, obtain the pixel offset value of the center point of the calibration plate relative to the standard value, obtain the offset through interpolation, and calculate the corrected offset in combination with the error compensation model.

[0136] Step 6.2: Compare whether the corrected offset is less than the required target accuracy range. If so, send a True instruction to the robot to make the end of the robotic arm 22 return to the standard position in Step 3.1 according to the corrected offset and drive the fixture 16 to complete the loading and unloading operation. Otherwise, send False to the actuator and perform Step (5) again to identify the current corrected offset value, and make the host computer send the corrected offset to the actuator for iterative adjustment.

[0137] Specifically, the target accuracy range is 0 to 0.1 mm. When the corrected offset is obtained as 0.03 mm, send a True instruction to the robot to make the end of the robotic arm 22 return to the standard position in Step 3.1 according to the offset value and drive the fixture 16 to complete the preset loading and unloading operation, and finally successfully and accurately complete the loading and unloading operation. The iterative method can be flexibly adapted according to the accuracy requirements of the on-site environment, avoid performance redundancy, and achieve scalable setting of the accuracy range.

[0138] Corresponding to the foregoing embodiments of the vision-based PCB board material autonomous precise positioning method, the present application also provides embodiments of a vision-based PCB board material autonomous precise positioning device.

[0139] Figure 8 It is a block diagram of a vision-based PCB board material autonomous precise positioning device shown according to an exemplary embodiment. Referring to Figure 8 , the device is applied to the processor of the robot and may include:

[0140] An acquisition module 31, configured to obtain a preset operating path of the robotic arm 22 after installing a camera support module 1 at the end of the robotic arm 22 and fixedly installing a camera 11 at the end of the L-shaped camera fixing member 15 in the camera support module 1 and the robotic arm 22 clamping the PCB board 17;

[0141] A solution module 32, configured to obtain a distorted calibration plate image captured under preset camera 11 parameters and solve the distortion coefficient according to the distorted calibration plate image;

[0142] A setting module 33, configured to obtain a standard position image of the positioning mark target 21 at the center of the camera 11 field of view, perform distortion correction on the standard position image using the distortion coefficient and then perform preprocessing to obtain an image area of the positioning mark target 21, extract feature points of the positioning mark target 21, calculate the coordinate value and vector value of the center pixel point of the mark target in the image coordinate system, and set the position of the end of the robotic arm 22 when the positioning mark target 21 is at the center of the camera 11 field of view as the origin in the coordinate system of the processing table 18;

[0143] A modeling module 34, configured to obtain a set of standard offset images acquired under preset camera 11 translation and rotation parameters, solve the image coordinate values of the center pixel points of the marking targets, record the correspondence between the preset standard offset and the image coordinate offset values of the center points of the marking targets, and establish an error compensation model;

[0144] A correction module 35, configured to obtain and solve the image coordinate values of the feature points of the image target to be solved, compare them with the center pixel difference of the positioning marking target 21 in the standard offset image set in step (4), interpolate between the two closest translation values and rotation values to obtain the actual offset, and drive the fixture 16 to correct the offset according to the actual offset;

[0145] A control module 36, configured to, after the fixture 16 corrects the offset, repeatedly execute the steps in the correction module in iteration until the difference between the recognized and positioned image of the marking target 21 and the standard position image is less than a preset target accuracy range, control the robotic arm 22 to return to the standard position according to the offset value, and drive the fixture 16 to complete the loading and unloading operation.

[0146] Regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0147] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial descriptions of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present application. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0148] Correspondingly, the present application further provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the vision-based PCB sheet material autonomous precise positioning method as described above. As Figure 9 shown, it is a hardware structure diagram of a device with any data processing ability where the deep learning dataset access system provided by the embodiment of the present invention is located. In addition to Figure 9 the processors, memory, and network interfaces shown, the device with any data processing ability where the device in the embodiment is located usually further includes other hardware according to the actual functions of the device with any data processing ability, which will not be elaborated herein.

[0149] Correspondingly, the present application also provides a computer-readable storage medium, on which computer instructions are stored. When the instructions are executed by a processor, the above-described vision-based autonomous and precise positioning method for PCB sheets is implemented. The computer-readable storage medium may be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium may also be an external storage device, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. equipped on the device. Further, the computer-readable storage medium may also include both an internal storage unit of any device with data processing capabilities and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and may also be used to temporarily store data that has been output or is to be output.

[0150] After considering the specification and practicing the content disclosed herein, those skilled in the art will readily conceive of other embodiments of the present application. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common knowledge or conventional technical means in the technical field not disclosed in the present application.

[0151] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A vision-based autonomous and precise positioning method for PCB sheet materials, characterized in that Including: Step (1): After installing a camera support module at the end of the robotic arm and fixedly installing a camera at the end of the L-shaped camera fixing part in the camera support module, and the robotic arm gripping the PCB board, obtain the pre-set operating path of the robotic arm; Step (2): Obtain the distorted calibration board image captured under the pre-set camera parameters, and solve the distortion coefficient according to the distorted calibration board image; Step (3): Obtain the standard position image with the positioning marker target at the center of the camera's field of view, perform distortion correction on the standard position image using the distortion coefficient and then perform pre-processing to obtain the positioning marker target image area, extract the feature points of the positioning marker target, calculate the coordinate value and vector value of the center pixel point of the marker target in the image coordinate system, and set the position of the end of the robotic arm when the positioning marker target is at the center of the camera's field of view as the origin in the processing table coordinate system; Step (4): Obtain the standard offset image set collected under the pre-set camera translation and rotation parameters, solve the image coordinate value of the center pixel point of the marker target, record the corresponding relationship between the pre-determined standard offset and the image coordinate offset value of the center point of the marker target, and establish an error compensation model; Step (5): Obtain and solve the image coordinate value of the feature point of the image target to be solved, compare it with the difference in the center pixel of the positioning marker target in the standard offset image set in step (4), interpolate between the two closest translation values and rotation values to obtain the actual offset, and drive the fixture to correct the offset according to the actual offset; Step (6): After the fixture corrects the offset, repeat step (5) iteratively until the difference between the recognized positioning marker target image and the standard position image is less than the pre-set target accuracy range, control the robotic arm to return to the standard position according to the offset value and drive the fixture to complete the loading and unloading operation.

2. The method according to claim 1, wherein In the said step (1), the process of "installing a camera support module at the end of the robotic arm and fixedly installing a camera at the end of the camera support, and the robotic arm gripping the PCB board" is as follows: Set a square connecting piece below the flange connecting shaft at the end of the robotic arm, fixedly connect it to the flange connecting shaft, install one side of the L-shaped camera fixing part on the flange connecting shaft, fixedly connect one side of the rib support piece to the flange connecting shaft through the square connecting piece, fixedly install the camera on the other side of the L-shaped camera fixing part, install a fixture directly below the flange connecting shaft and grip the PCB board; Wherein the axis of the flange connecting shaft coincides with the central axis of the robotic arm, the central axis of the PCB board deviates within 5 mm from the machine central axis, and the camera support deviates more than 280 mm from the machine central axis.

3. The method according to claim 1, characterized in that The said step (2) includes: Step 2.1: At the pre-set camera shooting height H, control the camera to shoot the distorted calibration board image, wherein the size of the distorted calibration board is larger than the shooting field of view range of the camera; Step 2.2: Extract the sub-pixel level feature points in the distorted calibration board image, and use the feature point image coordinate value and the actual coordinate value to calculate the distortion coefficient by combining radial and tangential distortions; Where k1, k2, k3, p1, and p2 are five distortion coefficients, x and y are the coordinates of the feature points in the distortion calibration plate image, and x correted ,y correted is the pixel coordinate of the corrected point.

4. The method according to claim 1, wherein The said step (3) includes: Step 3.1: Fix the positioning mark target on the processing table, make the vision center position of the positioning mark target under the fixed shooting point, and control the camera to take an image of the standard position under the fixed shooting point, where the positioning mark target is a small calibration board with an area less than 1 / 9 of the vision area within the shooting vision; Step 3.2: Distortion-correct the standard position image using the distortion coefficient, perform histogram normalization, median filtering, brightness correction, and mean binarization on the corrected image in sequence, and obtain the positioning mark target image area using area and perimeter features; Step 3.3: Solve the coordinates of all feature points in the image coordinate system and use the mean value of the coordinates of all feature points as the coordinates (x c , y c ) of the central pixel point of the positioning marker target: where (x pab , y pab ) are the coordinates of sub-pixel feature points in the image coordinate system, a and b represent the number of rows and columns of feature points in the positioning marker target, 1 ≤ a ≤ n, 1 ≤ b ≤ n; Step 3.4: Fit to obtain the vectors x1, …, x of each row of feature points in the image coordinate system n , and solve for the vector mean as the vector value of the central pixel point of the positioning marker target; Step 3.5: Set the position of the end of the robotic arm when the positioning mark target is at the exact center of the camera's vision as the origin in the processing table coordinate system.

5. The method according to claim 1, wherein The said step (4) includes: Step 4.1: Centering on the origin in the coordinate system of the processing table, control the actuator to perform a translation of ±c mm with the xy direction in the tool coordinate system of the robotic arm as the reference direction, and control the camera to take the first picture sets at the coordinates (-c, -c), (0, -c), (c, -c), (-c, 0), (0, 0), (c, 0), (-c, c), (0, c), and (c, c) respectively. Let the number of shooting times be n 2 times, the short side distance of the shooting field of view is l mm, where n ≥ 3 and l / 2n ≤ c ≤ l / n mm; Step 4.2: Centered at the origin in the processing table coordinate system, control the actuator to rotate ±d° with the z-axis in the tool coordinate system of the robotic arm as the rotation axis. At rotation angles of -nd°…-3d°…0°…3d°…n d°, control the camera to take the second set of pictures respectively. Assume the number of shooting times is 2n + 1, where n≥1 and 45 / (2n + 1)°≤d≤90 / (2n + 1)°; Step 4.3: Obtain the center point pixel coordinates and coordinate systems of the positioning mark target in the first set of pictures and the second set of pictures. According to the corresponding translation or rotation coordinate actual variables, obtain the translation vector and rotation vector of the offset of the processing table coordinate system and the pixel coordinate system; Step 4.4: Record the actual offset value of the robotic arm and the visually calculated offset value, collect the error data at each position, and establish an error compensation model.

6. The method according to claim 1, wherein The said step (5) includes: Step 5.1: Run the end of the robotic arm to the fixed shooting point to take a picture, and calculate the pixel coordinate system of the center point of the mark target; Step 5.2: Taking the pixel coordinate system of the center point of the mark target in the standard position as the reference value, obtain the pixel offset value of the center point of the mark target to be recognized relative to the standard value; Step 5.3: Take the two closest sets of pixel offset values, interpolate and solve for the translation amount and rotation amount between the two sets of data, and convert the pixel offset value into the actual offset amount; Step 5.4: According to the error compensation model, obtain the error value in the error compensation model corresponding to the pixel coordinates of the center point of the mark target, and correct the actual offset amount; Step 5.5: According to the corrected offset amount, drive the fixture to correct the offset.

7. The method according to claim 1, wherein The said step (6) includes: Step 6.1: Repeat step (5) to obtain the corrected offset amount; Step 6.2: Compare whether the corrected offset amount is less than the required target accuracy range. If so, send a True instruction to the robot, make the end of the robotic arm return to the standard position in step 3.1 according to the corrected offset amount, and drive the fixture to complete the loading and unloading action. Otherwise, send False to the actuator, perform step (5) again, identify the current corrected offset value, and make the upper computer send the corrected offset amount to the actuator for iterative adjustment.

8. An autonomous and precise positioning device for PCB sheets based on vision, characterized in that, Includes: An acquisition module, configured to obtain a pre-set robotic arm operation path after installing a camera support module at the end of a robotic arm and fixedly installing a camera at the end of an L-shaped camera fixing member in the camera support module, and after the robotic arm grips a PCB board; A solution module, configured to obtain a distorted calibration plate image captured under pre-set camera parameters, and solve for distortion coefficients according to the distorted calibration plate image; A setting module, configured to obtain a standard position image with a positioning mark target located at the exact center of the camera's field of view, perform preprocessing on the standard position image after distortion correction using the distortion coefficients to obtain a positioning mark target image area, extract feature points of the positioning mark target, calculate the coordinate value and vector value of the center pixel point of the mark target in the image coordinate system, and set the position of the end of the robotic arm when the positioning mark target is located at the exact center of the camera's field of view as the origin in the processing table coordinate system; A modeling module, configured to obtain a set of standard offset images collected under pre-set camera translation and rotation parameters, solve for the image coordinate value of the center pixel point of the mark target, record the correspondence between a pre-determined standard offset and the offset value of the image coordinates of the mark target center point, and establish an error compensation model; A correction module, configured to obtain and solve for the image coordinate value of the feature point of the image target to be solved, compare it with the difference in the center pixel of the positioning mark target in the standard offset image set in step (4), interpolate to obtain the actual offset between the two closest translation values and rotation values, and drive the fixture to correct the offset according to the actual offset; A control module, configured to, after the fixture corrects the offset, repeatedly execute the steps in the correction module iteratively until the difference between the recognized positioning mark target image and the standard position image is less than a pre-set target accuracy range, control the robotic arm to return to the standard position according to the offset value and drive the fixture to complete the loading and unloading operation; 9. An electronic device, characterized in that, Comprising: One or more processors; A memory, configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1-7.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that, When the instruction is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

Images