A 3D vision-based loading table position registration method

Abstract

The present application relates to a kind of based on 3D vision's loading platform position registration method, belong to position registration technical field, solve the positioning compensation of the precision of the loading platform in prior art is lower.The method of the present application includes: setting at least three circular marks on workpiece;After workpiece is transported to specified position by controlling loading platform, the image of workpiece in current position is collected as target image by 3D camera;Based on the ellipse fitting processing to circular mark, obtain the workpiece point cloud as target point cloud from target image;Target point cloud is matched with reference point cloud, and the translational deviation and rotational deviation between current position and specified position are obtained;Reference point cloud is the workpiece point cloud of specified position collected under template posture;Based on translational deviation and rotational deviation, control loading platform to move to specified position.The method of the present application can effectively improve the positioning compensation precision of loading platform.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of position registration, in particular to a 3D vision-based feeding table registration method. BACKGROUND

[0002] The feeding table position registration technology is mainly applied in industrial production to realize accurate positioning of parts to be processed. At present, there is a certain error in the positioning of the servo feeding table in industrial production. In order to ensure that the positioning accuracy of the workpiece to be processed meets the requirements of subsequent spraying and other processes after the feeding table is in place, the positioning error of the feeding table needs to be compensated.

[0003] 3D vision error positioning compensation is a method of detecting and correcting system error using three-dimensional vision information. The basic principle is to obtain the position information of the target object through a 3D camera, compare it with the expected position, detect the system error, and compensate by adjusting the system parameters to make the target object reach the expected position. Error positioning compensation using 3D vision has important value in industrial automation, robot control, aerospace, etc.

[0004] However, the positioning compensation accuracy of the existing 3D vision error positioning compensation technology is low, which cannot meet the accuracy requirements of the feeding table position registration, and is not conducive to the stability and adaptability of positioning. SUMMARY

[0005] In view of the above analysis, the embodiments of the present application aim to provide a 3D vision-based feeding table registration method to solve the problem of low positioning compensation accuracy of the feeding table in the prior art.

[0006] In one aspect, the embodiments of the present application provide a 3D vision-based feeding table position registration method, which comprises:

[0007] At least three circular marks are arranged on the workpiece;

[0008] After the feeding table transports the workpiece to the specified position, a 3D camera is used to collect an image of the workpiece at the current position as a target image;

[0009] Based on the elliptical fitting processing of the circular marks, the workpiece point cloud is obtained from the target image as a target point cloud;

[0010] The target point cloud is matched with the reference point cloud to obtain the translation deviation and rotation deviation between the current position and the specified position; the reference point cloud is the workpiece point cloud of the specified position collected under the template posture;

[0011] Based on the translation deviation and rotation deviation, the feeding table is controlled to move to the specified position.

[0012] Further improvement based on the above method, the method for obtaining workpiece point cloud from the target image as target point cloud based on the ellipse fitting processing of the circular mark includes the following steps:

[0013] Performing binaryzation processing on the target image to obtain a binaryzation image;

[0014] Based on the ellipse fitting processing of the circular mark on the binaryzation image, the three-dimensional coordinates of each circular mark in the image coordinate system are obtained;

[0015] Converting the coordinates of the circular mark in the image coordinate system into the coordinates of the loading table coordinate system;

[0016] Recording the obtained three-dimensional coordinates of the circular mark in the loading table coordinate system to form target point cloud data.

[0017] Further improvement based on the above method, performing binaryzation processing on the target image to obtain a binaryzation image includes:

[0018] For each pixel point in the target image, defining a neighborhood centered on the pixel point;

[0019] Calculating the local threshold value corresponding to the pixel point according to the pixel values of the neighborhood;

[0020] Comparing the current pixel value of each pixel point with its local threshold value, if the current pixel value is greater than the threshold value, setting the pixel value of the pixel point to the maximum value; if the pixel value of the current pixel point is less than the threshold value, setting the pixel value of the pixel point to zero, thereby obtaining a binaryzation image.

[0021] Further improvement based on the above method, the formula for calculating the local threshold value corresponding to the pixel point according to the pixel values of the neighborhood is:

[0022]

[0023]

[0024] In the formula, T(x, y) is the local threshold value corresponding to the pixel point (x, y), W(x, y) represents the neighborhood centered on the pixel point (x, y), I(i, j) is the pixel value corresponding to the pixel point (i, j), ω(i, j) is the weight corresponding to the pixel point (i, j), and σ is the standard deviation of the Gaussian distribution.

[0025] Further improvement based on the above method, the method for obtaining the three-dimensional coordinates of each circular mark in the image coordinate system based on the ellipse fitting processing of the circular mark on the binaryzation image includes the following steps:

[0026] An ellipse fitting is performed on the contour of the circular mark on the binary image to obtain an ellipse equation of each circular mark contour and coordinates of each ellipse center;

[0027] A plane fitting is performed according to the coordinates of each ellipse center to obtain a fitting plane equation and a normal vector of the fitting plane;

[0028] An affine transformation matrix is calculated according to the ellipse equation of the circular mark and the normal vector of the fitting plane;

[0029] The coordinates of each ellipse center are converted into corresponding center coordinates of the circular mark in the image coordinate system by the affine transformation matrix, and the center coordinates are taken as the three-dimensional coordinates of the circular mark in the image coordinate system.

[0030] Based on the further improvement of the above method, the plane fitting according to the coordinates of each circular mark in the image coordinate system to obtain a fitting plane equation and a normal vector of the fitting plane comprises:

[0031] Two circular marks are taken as a group, and a direction vector of a line connecting the center points of the two circular marks is calculated;

[0032] Based on the cross product of the direction vectors of the two lines, a least square method is used to solve the normal vector of the fitting plane and obtain the fitting plane equation.

[0033] Based on the further improvement of the above method, the calculation of the affine transformation matrix between the image coordinate system and the loading table coordinate system according to the ellipse equation of the circular mark and the normal vector of the fitting plane comprises the steps of:

[0034] The major axis a and the minor axis b of the ellipse are determined according to the ellipse equation of the circular mark;

[0035] The included angle θ between the normal vector of the fitting plane and the normal vector of the loading table plane is calculated;

[0036] The affine transformation matrix θ between the image coordinate system and the loading table coordinate system is calculated according to the major axis a and the minor axis b of the ellipse and the included angle θ, and the formula is T s :

[0037]

[0038] r = 0.5 × (a + b);

[0039] In the formula, r is the average radius of the ellipse.

[0040] Based on the further improvement of the above method, the target point cloud is matched with the reference point cloud based on the iterative closest point algorithm to obtain the translation deviation and the rotation deviation between the current position and the specified position.

[0041] Based on the further improvement of the above method, the target point cloud is matched with the reference point cloud based on the iterative closest point algorithm, and the translation deviation and the rotation deviation between the current position and the specified position are obtained, comprising the following steps:

[0042] Set the initial translation deviation and rotation deviation;

[0043] For each point in the target point cloud, find the closest point in the reference point cloud using the nearest neighbor search algorithm, and take the corresponding two points as a matching point pair;

[0044] According to the matching point pair, the error between the target point cloud and the reference point cloud is calculated;

[0045] Optimize the adjustment of the translation deviation and the rotation deviation to minimize the error between the target point cloud and the reference point cloud, so as to obtain the optimal translation deviation and rotation deviation as the translation deviation and rotation deviation of the current position and the specified position.

[0046] Based on the further improvement of the above method, the error between the target point cloud and the reference point cloud is calculated as follows:

[0047]

[0048] q i =Rp i +t;

[0049] In the formula, E(R,t) is the error, p i is the coordinate of the i-th point of the target point cloud, q i is the coordinate of the i-th point of the reference point cloud, R is the rotation deviation, and t is the translation deviation.

[0050] Compared with the prior art, the present application can at least achieve one of the following beneficial effects:

[0051] 1. The present application provides a 3D vision-based loading table position registration method, which can effectively improve the accuracy, stability and adaptability of positioning compensation, realize automatic positioning compensation of the workpiece to be processed, and effectively reduce the labor intensity and improve the production efficiency.

[0052] 2、In the present application, by setting a plurality of circular marks distributed along the circumference on the workpiece, and using a 3D camera to collect images, and combining with elliptical fitting processing to obtain the current workpiece point cloud, i.e. target point cloud, the precise positioning of the workpiece can be realized, then the target point cloud is matched with the reference point cloud to obtain the translation deviation and rotation deviation between the current position and the specified position, and the control loading table is moved to the specified position based on the translation deviation and rotation deviation, such processing can automatically correct the position and attitude of the workpiece, improve the assembly or machining precision, and reduce manual intervention, improve the work efficiency and automation level.

[0053] The above technical solutions can be combined with each other in the present application to realize more preferred combination solutions. Other features and advantages of the present application will be described in the subsequent specification, and some advantages will become apparent from the specification, or will be understood by implementing the present application. The purposes and other advantages of the present application can be realized and obtained from the contents specifically pointed out in the specification and the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0054] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated in and constitute a part of this application, illustrate embodiments of the application and are used to explain the principles of the application, but are not used to limit the scope of the application.

[0055] Figure 1 A flow chart of a workpiece position registration method based on 3D vision according to an embodiment of the present application;

[0056] Figure 2 A schematic diagram of the distribution position of circular marks on a workpiece according to an embodiment of the present application;

[0057] Reference signs:

[0058] 1-workpiece; 2-circular mark. DETAILED DESCRIPTION

[0059] The preferred embodiments of the present application will be specifically described below in conjunction with the drawings, wherein the drawings constitute a part of this application, and are used to explain the principles of the embodiments of the present application, but are not used to limit the scope of the present application.

[0060] An embodiment of the present application provides a workpiece position registration method based on 3D vision, as shown in Figure 1 The method comprises the following steps:

[0061] Step 1: setting at least three circular marks 2 on the workpiece 1;

[0062] Step 2: after the workpiece 1 is transported to the specified position by the control loading table, the image of the workpiece 1 at the current position is collected by the 3D camera as a target image;

[0063] Step 3, based on the elliptical fitting processing of the circular mark 2, the workpiece point cloud is obtained from the target image as a target point cloud;

[0064] Step 4, the target point cloud is matched with the reference point cloud to obtain the translation deviation and the rotation deviation between the current position and the specified position; the reference point cloud is the workpiece point cloud of the specified position collected under the template posture;

[0065] Based on the translation deviation and the rotation deviation, the feeding table is controlled to move to the specified position.

[0066] Compared with the prior art, the embodiment of the present application provides a feeding table position registration method based on 3D vision, which can effectively improve the accuracy, stability and adaptability of positioning compensation, realize automatic positioning compensation of the workpiece 1 to be processed, and effectively reduce the labor intensity and improve the production efficiency.

[0067] Specifically, in the embodiment of the present application, a plurality of circular marks 2 are arranged on the workpiece 1 along the circumference, and the image is collected by using a 3D camera, and the current workpiece point cloud, i.e. the target point cloud, is obtained by combining the elliptical fitting processing, which can realize accurate positioning of the workpiece 1. Then, the target point cloud is matched with the reference point cloud to obtain the coordinate transformation matrix, which is decomposed into translation deviation and rotation deviation, and the feeding table is controlled to move to the specified position based on the translation deviation and the rotation deviation. Such processing can automatically correct the position and attitude of the workpiece 1, improve the assembly or machining accuracy, reduce manual intervention, improve the work efficiency and automation level.

[0068] Step 1, at least three circular marks 2 are arranged on the workpiece 1.

[0069] When implemented, at least three circular marks 2 are arranged on the workpiece 1 along the circumference, as shown in Figure 2 .

[0070] Step 2, after the feeding table transports the workpiece 1 to the specified position, the image of the workpiece 1 at the current position is collected by the 3D camera as a target image.

[0071] It should be noted that the feeding table of the embodiment of the present application transports the workpiece 1 through translation and rotation movements. After receiving the control instruction, the feeding table transports the workpiece 1 to the specified position, but due to the positioning deviation, the feeding table cannot accurately stop at the specified position, and thus the positioning of the workpiece 1 is deviated, i.e. there is a deviation between the current position of the workpiece 1 and the specified position.

[0072] The 3D camera can obtain a planar image and depth information of the photographed object, i.e. three-dimensional position information.

[0073] Step 3, based on the elliptical fitting processing of the circular mark 2, the workpiece point cloud is obtained from the target image as a target point cloud, including the following steps:

[0074] The target image is binarized to obtain a binarized image;

[0075] Based on the elliptical fitting processing of the circular mark 2 on the binarized image, the three-dimensional coordinates of each circular mark 2 in the image coordinate system are obtained;

[0076] The coordinates of the circular mark 2 in the image coordinate system are converted into the coordinates of the loading table coordinate system;

[0077] The obtained three-dimensional coordinates of the circular mark 2 in the loading table coordinate system are recorded to form target point cloud data.

[0078] In the embodiment of the application, high-precision positioning of the circular mark 2 can be achieved through binarization processing and elliptical fitting. The binarization processing simplifies the gray value of the image to two values, usually 0 and 255. The binarization processing has the effects of simplifying the image, enhancing the contrast, and reducing noise, which facilitates subsequent elliptical fitting and edge detection.

[0079] Specifically, the target image is binarized to obtain a binarized image, including:

[0080] For each pixel point in the target image, a neighborhood is defined with the pixel point as the center;

[0081] The local threshold value corresponding to the pixel point is calculated according to the pixel values of the neighborhood;

[0082] The current pixel value of each pixel point is compared with its local threshold value. If the current pixel value is greater than the threshold value, the pixel value of the pixel point is set to the maximum value. If the pixel value of the current pixel point is less than the threshold value, the pixel value of the pixel point is set to zero, thereby obtaining a binarized image.

[0083] In the embodiment of the application, when performing binarization processing, an adaptive threshold processing technique is used to complete threshold processing of the image using a variable threshold. The adaptive threshold processing adaptively determines the threshold for different local regions, the result is more accurate, the image segmentation accuracy is improved, and the details of the image can be better preserved.

[0084] Specifically, the formula for calculating the local threshold value corresponding to the pixel point according to the pixel values of the neighborhood is:

[0085]

[0086]

[0087] In the formula, T(x, y) is a local threshold corresponding to a pixel point (x, y), W(x, y) represents a neighborhood with the pixel point (x, y) as a center, I(i, j) is a pixel value corresponding to a pixel point (i, j), ω(i, j) is a weight corresponding to the pixel point (i, j), and σ is a standard deviation of a Gaussian distribution.

[0088] In the embodiment of the application, a local threshold is obtained by calculating a weighted average value of a nearby region around each pixel point, and the threshold is used to process the current pixel point.

[0089] After the binarization processing, the image only contains two kinds of pixel values, the complexity of the image is reduced, and the contour and shape features in the image are highlighted, so that the accuracy of subsequent ellipse fitting can be improved.

[0090] Specifically, the ellipse fitting processing on the circular markers 2 on the binarized image is based on the following steps:

[0091] The contour of the circular marker 2 on the binarized image is subjected to ellipse fitting to obtain an ellipse equation of each circular marker 2 contour and a coordinate of each ellipse center;

[0092] A fitting plane equation and a normal vector of the fitting plane are obtained by plane fitting according to the coordinate of each ellipse center;

[0093] An affine transformation matrix is calculated according to the ellipse equation of the circular marker 2 and the normal vector of the fitting plane;

[0094] The coordinate of each ellipse center is converted into a corresponding center coordinate of a circle through the affine transformation matrix, and the center coordinate of the circle is taken as a three-dimensional coordinate of the circular marker 2 in an image coordinate system.

[0095] Further specifically, the contour of the circular marker 2 on the binarized image is subjected to ellipse fitting to obtain an ellipse equation of each circular marker 2 contour and a coordinate of each ellipse center, including the following steps:

[0096] An edge point coordinate of the circular marker 2 on the binarized image is extracted by using an edge detection algorithm, and a least square method is used to fit the edge point coordinate to obtain an ellipse equation of the circular marker 2 contour;

[0097] The coordinate of the ellipse center is obtained based on the ellipse equation of the circular marker 2 contour.

[0098] The standard equation of the ellipse is:

[0099] Ax 2 +bxy+Cy 2 +Dx+Ey+F=0;

[0100] The quadratic form of the ellipse standard equation is:

[0101] x T Qx=0;

[0102]

[0103]

[0104] The target function for fitting the edge point coordinates by using the least square method is:

[0105] X opt =argmin(E ellipse );

[0106]

[0107] In the formula, X opt is an optimization target, E ellipse is an error; u i and v i are respectively coordinates of an i-th edge point of the extracted center mark of the circle; u projected and v projected are respectively coordinates of the i-th edge point to be solved.

[0108] Further specifically, the fitting plane equation and the normal vector of the fitting plane are obtained according to the coordinates of each circular mark 2 in the image coordinate system, comprising:

[0109] Two circular marks 2 are taken as a group, and a direction vector of a line connecting the centers of the two circular marks 2 is calculated;

[0110] Based on the cross product of the direction vectors of the two lines, a normal vector of the fitting plane is solved by using the least square method, and a fitting plane equation is obtained.

[0111] The fitting plane equation is: n x X+n y Y+n z Z=0;

[0112] The plane normal vector is:

[0113] Wherein, X, Y, Z are respectively coordinates of an X axis, a Y axis and a Z axis in an image coordinate system, n x , n y , n z are respectively components of the normal vector of the fitting plane in the X axis, the Y axis and the Z axis.

[0114] The normal vector of the fitting plane can be obtained through the direction vectors of any two lines The cross product of the two vectors gives the normal vector of the plane, and the formula is:

[0115] Since the direction vectors of any two connecting lines can be calculated, the least square method is used for fitting to obtain the optimal fitting plane normal vector value.

[0116] Further specifically, the affine transformation matrix between the image coordinate system and the loading table coordinate system is calculated according to the ellipse equation of the circular mark 2 and the normal vector of the fitting plane, including the steps of:

[0117] According to the ellipse equation of the circular mark 2, the major axis a and the minor axis b of the ellipse are determined;

[0118] The angle θ between the normal vector of the fitting plane and the normal vector of the loading table plane is calculated;

[0119] According to the major axis a and the minor axis b of the ellipse and the angle θ, the affine transformation matrix θ between the image coordinate system and the loading table coordinate system is calculated, and the formula is T s :

[0120]

[0121] r = 0.5 * (a + b);

[0122] In the formula, r is the average radius of the ellipse.

[0123] It should be noted that the normal vector of the loading table plane is known,

[0124] Specifically, the coordinates of the circular mark 2 in the image coordinate system are converted into the coordinates of the loading table coordinate system, including:

[0125] According to the intrinsic matrix of the 3D camera, the coordinates of the circular mark 2 in the image coordinate system are converted into the coordinates of the camera coordinate system;

[0126] According to the extrinsic matrix of the 3D camera, the coordinates of the circular mark 2 in the image coordinate system are converted into the coordinates of the loading table coordinate system.

[0127] Wherein, according to the intrinsic matrix of the 3D camera, the coordinates of the circular mark 2 in the image coordinate system are converted into the coordinates of the camera coordinate system, and the formula is as follows:

[0128]

[0129] In the formula, [u v 1] T is the coordinate expression of the circular mark 2 in the image coordinate system, [x c y c z c ]T is the coordinate expression of the circular mark 2 in the camera coordinate system, and K is an intrinsic matrix.

[0130] The company of the intrinsic matrix is:

[0131] According to the extrinsic matrix of the 3D camera, the coordinates of the circular mark 2 in the image coordinate system are converted into the coordinates in the loading table coordinate system, and the formula is:

[0132]

[0133] In the formula, left [x w y w z w 1] T is the coordinate point expression of the circular mark 2 in the loading table coordinate system, [x c y c z c 1] T is the elliptical coordinate point expression in the camera coordinate system, and T is an extrinsic matrix.

[0134] The formula of the extrinsic matrix is:

[0135]

[0136] Wherein, R is a 3*3 rotation matrix, and t is a 3*1 translation matrix.

[0137] In implementation, the intrinsic matrix and the extrinsic matrix of the 3D camera can be obtained through calibration.

[0138] Specifically, the target point cloud and the reference point cloud are matched based on the iterative closest point algorithm to obtain the translation deviation and the rotation deviation between the current position and the specified position.

[0139] Further specifically, the target point cloud and the reference point cloud are matched based on the iterative closest point algorithm to obtain the translation deviation and the rotation deviation between the current position and the specified position, including the following steps:

[0140] Set the initial translation deviation and the rotation deviation;

[0141] For each point in the target point cloud, the nearest neighbor search algorithm is used to find the nearest point in the reference point cloud, and the corresponding two points are taken as a matching point pair;

[0142] According to the matching point pair, the error between the target point cloud and the reference point cloud is calculated;

[0143] The translation deviation and the rotation deviation are optimized to minimize the error between the target point cloud and the reference point cloud, so as to obtain an optimal translation deviation and rotation deviation as the translation deviation and rotation deviation of the current position and the specified position.

[0144] The reference point cloud is defined as P=(p1, p2, p3,..., pN), where N is the number of points in the reference point cloud. N The coordinates of each point in the reference point cloud are p i =(x i , y i , z i ) T . The target point cloud is Q=(q1, q2, q3,..., qN), where N is the number of points in the target point cloud. N The coordinates of each point in the target point cloud are q i =(x i ', y i ', z i ') T .

[0145] The error between the target point cloud and the reference point cloud is calculated as follows:

[0146]

[0147] q i = Rp i + t

[0148] where E(R, t) is the error, p i is the coordinates of the i-th point in the reference point cloud, q i is the coordinates of the i-th point in the target point cloud, R is the rotation deviation, and t is the translation deviation.

[0149] In implementation, the least squares method is used to optimize and adjust the translation deviation and the rotation deviation.

[0150] The formula of the translation deviation and the rotation deviation is as follows:

[0151] R = VU T ;

[0152]

[0153] where U and V T are the left singular vector matrix and the right singular vector matrix of the covariance matrix H, respectively, and are the centroids of the reference point cloud and the target point cloud, respectively. The centroids are calculated as follows:

[0154] Calculate the centroids of the reference point cloud and the target point cloud and

[0155]

[0156]

[0157] Centralized data:

[0158]

[0159]

[0160] Compute the covariance matrix H:

[0161]

[0162] Singular value decomposition:

[0163] H = U∑V T ;

[0164] Singular value decomposition (SVD) is performed on the covariance matrix H to obtain the left singular vector matrix U, the singular value matrix Σ, and the right singular vector matrix V T .

[0165] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiment methods can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer readable storage medium. The computer readable storage medium is a disk, an optical disk, a read-only memory, a random access memory, etc.

[0166] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed by the present application can be easily thought of by those skilled in the art, and should be covered within the protection scope of the present application.

Claims

1. A 3D vision-based loading table position registration method, characterized in that, The method comprises: Setting at least three circular marks on a workpiece; After the feeding table transports the workpiece to the designated position, acquiring an image of the workpiece at the current position as a target image by a 3D camera; Based on the ellipse fitting processing of the circular marks, obtaining a workpiece point cloud from the target image as a target point cloud; Based on the iterative closest point algorithm, matching the target point cloud with a reference point cloud to obtain the translation deviation and the rotation deviation between the current position and the designated position, comprising the following steps: Setting an initial translation deviation and an initial rotation deviation; For each point in the target point cloud, finding the closest point in the reference point cloud using the nearest neighbor search algorithm, and taking the corresponding two points as a matching point pair; According to the matching point pair, calculating the error between the target point cloud and the reference point cloud: q i = Rp i + t; where E(R, t) is the error, p i is the coordinate of the i-th point of the target point cloud, q i is the coordinate of the i-th point of the reference point cloud, R is the rotation bias, and t is the translation bias. Optimizing and adjusting the translation deviation and the rotation deviation to minimize the error between the target point cloud and the reference point cloud, thereby obtaining the optimal translation deviation and the optimal rotation deviation as the translation deviation and the rotation deviation between the current position and the designated position; The reference point cloud is a workpiece point cloud of the designated position acquired under a template posture; Based on the translation deviation and the rotation deviation, controlling the feeding table to move to the designated position.

2. The load table position registration method of claim 1, wherein, Based on the ellipse fitting processing of the circular marks, obtaining a workpiece point cloud from the target image as a target point cloud, comprising the following steps: Performing binaryzation processing on the target image to obtain a binaryzation image; Based on the ellipse fitting processing of the circular marks on the binaryzation image, obtaining the three-dimensional coordinates of each circular mark in the image coordinate system; Converting the coordinates of the circular marks in the image coordinate system into coordinates in the feeding table coordinate system; Recording the obtained three-dimensional coordinates of the circular marks in the feeding table coordinate system to form target point cloud data.

3. The load table position registration method of claim 2, wherein, Performing binaryzation processing on the target image to obtain a binaryzation image, comprising: For each pixel point in the target image, defining a neighborhood centered on the pixel point; Calculating the local threshold value corresponding to the pixel point according to the pixel values of the neighborhood; Comparing the current pixel value of each pixel point with its local threshold value, if the current pixel value is greater than the threshold value, setting the pixel value of the pixel point to the maximum value, if the current pixel value of the pixel point is less than the threshold value, setting the pixel value of the pixel point to zero, thereby obtaining a binaryzation image.

4. The load table position registration method of claim 3, wherein, The formula for calculating the local threshold value corresponding to the pixel point according to the pixel values of the neighborhood is: In the formula, T(x, y) is the local threshold value corresponding to the pixel point (x, y), W(x, y) represents the neighborhood centered on the pixel point (x, y), I(i, j) is the pixel value corresponding to the pixel point (i, j), ω(i, j) is the weight corresponding to the pixel point (i, j), and σ is the standard deviation of the Gaussian distribution.

5. The load table position registration method of claim 2, wherein, The ellipse fitting processing of the circular marks on the binaryzation image to obtain the three-dimensional coordinates of each circular mark in the image coordinate system comprises the following steps: Performing ellipse fitting on the contours of the circular marks on the binaryzation image to obtain the ellipse equation of each circular mark contour and the coordinates of each ellipse center; According to the coordinates of each ellipse center, a plane fitting is performed to obtain a fitting plane equation and a normal vector of the fitting plane; An affine transformation matrix is calculated according to the ellipse equation of the circular mark and the normal vector of the fitting plane; The coordinates of each ellipse center are converted into corresponding circle center coordinates by the affine transformation matrix, and the circle center coordinates are taken as the circular mark in the three-dimensional coordinates of the image coordinate system.

6. The load table position registration method of claim 5, wherein, According to the coordinates of each circular mark in the image coordinate system, a plane fitting is performed to obtain a fitting plane equation and a normal vector of the fitting plane, including: Two circular marks are taken as a group, and a direction vector of a line connecting the centers of the two circular marks is calculated; Based on the cross product of the direction vectors of the two lines, a normal vector of the fitting plane is solved by using a least square method, and a fitting plane equation is obtained.

7. The method of claim 5, wherein, An affine transformation matrix between the image coordinate system and the loading table coordinate system is calculated according to the ellipse equation of the circular mark and the normal vector of the fitting plane, including steps of: According to the ellipse equation of the circular mark, a major axis a and a minor axis b of the ellipse are determined; An included angle θ between the normal vector of the fitting plane and the normal vector of the loading table plane is calculated; An affine transformation matrix T between the image coordinate system and the loading table coordinate system is calculated according to the major axis a and the minor axis b of the ellipse and the included angle θ, and the formula is T s : r = 0.5 × (a + b); In the formula, r is the average radius of the ellipse.

Images