A rotating disc surface circle center calibration method based on ellipse fitting
Patent Information
- Application Number
- CN202611029905.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-10
- Publication Date
- 2026-09-25
AI Technical Summary
若简单对轨迹数据求取几何中心而不考虑倾斜对轨迹形状的影响，得到的几何中心并非真实旋转轴在图像中的投影位置，二者之间存在系统性偏差；现有技术对此缺乏自动化的倾斜识别与修正手段，导致标定结果在倾斜工况下不可靠
[0009]本发明与现有技术相比，具有以下实质性改进：
Smart Images
Figure CN122820801A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of visual calibration technology for industrial robot rotary tables, specifically to a method for calibrating the center of a rotating disk based on ellipse fitting. Background Technology
[0002] In industrial robot docking and other operations involving rotating worktables, it is necessary to accurately calibrate the position of the center of the rotating disk in the camera image coordinate system. This serves as the reference point for subsequent workpiece positioning and angle calculation. Existing technical solutions have the following shortcomings: Firstly, the traditional method involves manually marking the center position, requiring operators to manually select the center coordinates in the image. Since different operators have different judgment standards, the same calibration will yield different results among different operators, leading to poor calibration consistency. Furthermore, each manual marking process is time-consuming, requiring repeated calibration when the production line frequently changes workpiece types, severely slowing down the production cycle.
[0003] Secondly, some automatic calibration solutions require precise encoder angle feedback from the rotating mechanism. This means that the current rotation angle value must be acquired synchronously with each image acquisition, and then the center of rotation must be calculated based on the angle-coordinate correspondence. However, many rotary tables used in industrial settings are not precision angle measuring devices and do not have high-precision angle sensors. They can only perform rotational movements but cannot output angle readings, making it impossible to directly deploy these automatic calibration solutions that rely on angle feedback.
[0004] Third, when the rotating disk is tilted, meaning there is a non-zero angle between the disk's normal and the camera's optical axis, the trajectory formed by the fixed markers on the disk as it rotates is not a perfect circle but an ellipse in the image. If the geometric center is simply calculated from the trajectory data without considering the influence of tilt on the trajectory shape, the obtained geometric center is not the projection position of the true rotation axis in the image, resulting in a systematic deviation between the two. Existing technologies lack automated tilt recognition and correction methods, leading to unreliable calibration results under tilted conditions. Summary of the Invention
[0005] The purpose of this invention is to provide a method for calibrating the center of a rotating disk based on ellipse fitting. This method only requires acquiring pixel images of fixed marker points on the rotating disk at different angles, automatically calculating the precise projection position of the rotation axis in the image coordinate system using an ellipse fitting algorithm, and evaluating the degree of disk tilt based on the ellipse axis ratio. Calibration can be completed without the need for angle encoder feedback.
[0006] Technical solution
[0007] A method for calibrating the center of a rotating disk based on ellipse fitting includes at least the following steps: Step 1: Drive the rotating disk to a specified number of different angular positions, and acquire an image containing a fixed marker point at each angular position. The marker point is a specified fixed physical point on the rotating disk. Step 2: Extract the pixel coordinates of the fixed marker points for each acquired image to obtain a set of pixel coordinate points corresponding to each angle position. The pixel coordinates include row coordinates and column coordinates. Step 3: Use the set of pixel coordinate points as the input point set for ellipse fitting, and use the least squares ellipse fitting algorithm to calculate the parameters of the ellipse equation, including the coordinates of the ellipse center, the length of the major axis, the length of the minor axis, and the direction angle of the major axis; Step 4: Using the center of the ellipse obtained from ellipse fitting as the projection position of the rotation axis in the image coordinate system, and evaluating the tilt of the rotating disk surface by the ratio of the minor axis to the major axis. This method is executed by the robot's end-side processor.
[0008] Beneficial effects
[0009] Compared with the prior art, the present invention has the following substantial improvements: Firstly, addressing the issues of low efficiency and poor consistency in manual marking, this invention automatically acquires multi-angle images and directly calculates the center coordinates using an ellipse fitting algorithm. This eliminates subjective biases introduced by manual operation, and the calibration results depend solely on the algorithm's processing of the acquired data, resulting in high repeatability. In actual tests, automatic calibration takes 3 to 5 seconds, while manual marking takes an average of over 30 seconds, improving efficiency by 6 to 10 times. The standard deviation of the repeatability calibration error is less than 0.3 pixels, with consistency far superior to the fluctuations of over 1.5 pixels in manual marking. Changing image acquisition and coordinate calculation to a fully automated process is not simply a matter of automation; the key lies in the invention's discovery that the rotation trajectory of fixed marker points on the disk forms an ellipse rather than a perfect circle under tilted conditions. Therefore, ellipse fitting, rather than circular fitting, is used to obtain the center, a choice that is impossible to achieve in traditional methods relying on manual marking.
[0010] Secondly, addressing the issue of existing automated solutions requiring angle encoder feedback, this invention does not rely on any angle sensor data. Its core principle lies in the fact that the ellipse fitting algorithm itself only needs a set of pixel coordinates of the marked points, without needing to know the rotation angle value corresponding to each point. This is because the parameter estimation of the ellipse equation is based on the geometric distribution of the point set rather than the angular order; the spatial distribution information of each point in the set is sufficient to determine the ellipse parameters. Experimental verification shows that only five or more sampling points at different angles are needed to reduce the ellipse fitting residual to below 0.5 pixels, without requiring any angle readings. This discovery allows the invention to be directly deployed on ordinary rotary tables without angle sensors, lowering the hardware threshold and saving 2000 to 5000 yuan in hardware modification costs per unit.
[0011] Thirdly, addressing the issues of trajectory distortion caused by disk tilt and the lack of automatic tilt correction in existing technologies, this invention quantitatively assesses the degree of disk tilt using the ratio of the minor axis to the major axis obtained from ellipse fitting, and uses the center of the ellipse, rather than the geometric center, as the projection position of the rotation axis. Under tilt conditions, the center of the fitted ellipse precisely corresponds to the projection point of the rotation axis on the image plane, and this center will not drift due to changes in the tilt direction or angle. In actual measurements, when the disk tilt angle is within the range of 0° to 30°, the deviation of the calibrated center by this invention does not exceed 0.8 pixels, while the deviation of the method of directly taking the geometric center can reach more than 5 pixels as the tilt angle increases. This characteristic is based on the geometric principle that the perspective projection of a spatial circle is an ellipse, rather than a simple empirical correction, thus ensuring the reliability of the calibration results under tilt conditions and providing a quantitative basis for subsequent tilt compensation. Attached Figure Description
[0012] Figure 1 This is a schematic diagram of the overall process in an embodiment of the present invention.
[0013] Figure 2 This is a schematic diagram of circular motion data acquisition in an embodiment of the present invention.
[0014] Figure 3 This is a schematic diagram of the ellipse fitting results in an embodiment of the present invention.
[0015] The meanings of the markings in the diagram are as follows: 100 - Multi-angle image acquisition step, 101 - Marker point pixel coordinate extraction step, 102 - Ellipse fitting calculation step, 103 - Circle center determination step, 104 - Tilt evaluation step. 200 - Rotating disk area; 201 - Sampling positions of marker points at different angles; 300 - The fitted ellipse, 301 - The center of the ellipse, i.e., the projection position of the rotation axis, 302 - The major axis of the ellipse, 303 - The minor axis of the ellipse. Detailed Implementation
[0016] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0017] This invention runs on the end-side processor of an industrial robot and utilizes a robot vision system to automatically calibrate the rotating disk surface. The camera resolution is 640×480 pixels. The specific implementation steps are as follows: like Figure 1As shown by marker 100, the robot's end-side processor controls the rotating disk to rotate intermittently within a specified angular range, from 0 degrees to 360 degrees, triggering the camera to acquire images at a specified number of different angular positions. In this embodiment, the sampling angular positions are set to 8 groups, with non-uniformly distributed angular intervals of 0°, 52°, 97°, 143°, 195°, 248°, 301°, and 347°. At each angular position, a grayscale image containing fixed marker points is acquired, such as... Figure 2 On the rotating disk area indicated by mark 200, mark 201 indicates the sampling positions of the marker points at different angles. The fixed marker points are pre-drawn red dots on the rotating disk surface, with a physical diameter of 5 mm. At least 5 angle positions are used to ensure the reliability of the ellipse fitting.
[0018] like Figure 1 As shown in label 101, for each image acquired in step one, the processor first executes the image preprocessing pipeline, including distortion correction, boundary scaling, and circular ROI cropping operations, to obtain the image after ROI cropping. Distortion correction uses the camera intrinsic parameter matrix and distortion coefficients obtained by Zhang Zhengyou calibration method. The boundary scaling ratio is 5% of the image width and height, i.e., 32 pixels are cropped from the edge. The circular ROI has the image center as the center and a radius of 280 pixels. Then, color filtering is performed on the RGB three-channel image. Under the specified color threshold condition, i.e., the BGR channel value is within a tolerance range of ±15, the RGB values of the red marker point are matched to 180, 40, 40. The pixel region of the marker point is extracted, and the mean of the row coordinates and the mean of the column coordinates of all pixels in the region are calculated as the centroid pixel coordinates of the marker point. The pixel coordinates corresponding to all 8 angular positions are collected into a coordinate point set.
[0019] like Figure 1 As shown in label 102, the processor uses a least-squares ellipse fitting function to fit an ellipse to the set of coordinate points. This function takes eight points as input in the format of column coordinates of pixel coordinates as the horizontal axis and row coordinates as the vertical axis, and outputs the center coordinates, axis lengths, and orientation angle parameters of the fitted ellipse equation, as shown below. Figure 3 The fitted ellipse is shown in the center, marked 300. Since the input point set comes from the projection of the same fixed physical point on the rotating disk at different rotation angles, these points should fall on a perfect circle on an ideal horizontal disk, but on an ellipse on an inclined disk. Therefore, the major axis direction angle of the fitted ellipse reveals the direction of inclination, as shown... Figure 3 The major axis of the ellipse is shown at mark 302, and the minor axis is shown at mark 303. The ratio of the lengths of the major and minor axes reveals the degree of inclination. The ellipse fitting was performed using a least-squares method based on algebraic distance, with a fitting error threshold of 1.0 pixel.
[0020] like Figure 1As shown in marks 103 and 104, the processor determines the coordinates of the ellipse center in the ellipse fitting result as the projection position of the rotation axis in the ROI image coordinate system, as follows: Figure 3 The center of the ellipse marked 301 is the projection position of the rotation axis, recorded as the center coordinates in the row and column directions, respectively. Simultaneously, the ratio of the minor axis to the major axis is calculated. If this ratio is greater than 0.95, the disk surface is nearly horizontal; if the ratio is less than 0.95, the disk surface is tilted, requiring correction to the subsequent pixel-coordinate mapping based on the tilt parameters. In subsequent workpiece pose calculations, this rotation axis projection position is used as the pixel origin reference position for each workpiece category. Combined with the pre-calibrated pixel-to-real-coordinate mapping matrix, the pixel offset can be converted into the actual offset in the robotic arm coordinate system. The pixel-to-real-coordinate mapping matrix is obtained through checkerboard calibration with a calibration accuracy of 0.1 mm / pixel.
Claims
1. A method for calibrating the center of a rotating disk based on ellipse fitting, characterized in that, The method includes at least the following steps: Step 1: Drive the rotating disk to rotate to a specified number of different angular positions and acquire images containing fixed marker points, where each marker point is a specified fixed physical point on the rotating disk; Step 2: Extract the pixel coordinates of the fixed marker points from each acquired image to obtain a set of pixel coordinate points corresponding to each angular position, where the pixel coordinates include row coordinates and column coordinates; Step 3: Use the set of pixel coordinate points as the input point set for ellipse fitting, and calculate the parameters of the ellipse equation using a least-squares ellipse fitting algorithm, including the coordinates of the ellipse center, the length of the major axis, the length of the minor axis, and the direction angle of the major axis; Step 4: Use the center of the ellipse obtained by ellipse fitting as the projection position of the rotation axis in the image coordinate system, and evaluate the tilt of the rotating disk using the ratio of the minor axis to the major axis. The method is executed by a robot end-side processor.
2. The method according to claim 1, characterized in that, The fixed marker points mentioned in step one are designated fixed physical feature points on the rotating disk. The number of angle positions is at least five to ensure the reliability of ellipse fitting. The angle distribution is non-uniform and arbitrary. The image acquisition operation is controlled and executed by the robot end-side processor.
3. The method according to claim 1, characterized in that, Step 2, the pixel coordinate extraction, includes performing distortion correction, boundary scaling, and circular ROI cropping preprocessing on the image, then filtering and extracting the marker pixel region through color channel thresholding, and calculating the mean row coordinates and mean column coordinates of the pixels within the region as the centroid coordinates of the marker point.
4. The method according to claim 1, characterized in that, The ellipse fitting in step three uses the least squares ellipse fitting algorithm. The input point set is a data pair with column coordinates as the horizontal axis and row coordinates as the vertical axis. The output fitting ellipse parameters include the center column coordinates, the center row coordinates, the length of the major axis, the length of the minor axis, and the direction angle of the major axis.
5. The method according to claim 1, characterized in that, The ellipse center mentioned in step four is used as the pixel origin reference position for subsequent workpiece pose calculation. Combined with the specified pixel-to-real coordinate mapping matrix, the pixel offset is converted into the real offset in the robot arm coordinate system.
6. A rotating disk center calibration system based on ellipse fitting, characterized in that, include: A first module is used to drive the rotating disk to rotate to a specified number of different angular positions and acquire images containing fixed marker points, where each marker point is a specified fixed physical point on the rotating disk. A second module is used to extract the pixel coordinates of the fixed marker points from each acquired image to obtain a set of pixel coordinate points corresponding to each angular position, where the pixel coordinates include row coordinates and column coordinates. A third module is used to use the set of pixel coordinate points as the input point set for ellipse fitting and to calculate the parameters of the ellipse equation using a least-squares ellipse fitting algorithm, where the parameters include the coordinates of the ellipse center, the length of the major axis, the length of the minor axis, and the direction angle of the major axis. A fourth module is used to use the center of the ellipse obtained by ellipse fitting as the projection position of the rotation axis in the image coordinate system and to evaluate the degree of tilt of the rotating disk by the ratio of the minor axis to the major axis.
7. The system according to claim 6, characterized in that, The pixel coordinate extraction described in the second module includes performing distortion correction, boundary scaling, and circular ROI cropping preprocessing on the image, then filtering and extracting the pixel region of the marker point through color channel thresholding, and calculating the mean row coordinates and mean column coordinates of the pixels in the region as the centroid coordinates of the marker point.
8. The system according to claim 6, characterized in that, The ellipse fitting described in the third module adopts the least squares ellipse fitting algorithm. The input point set is a data pair with column coordinates as the horizontal axis and row coordinates as the vertical axis. The output fitting ellipse parameters include the center column coordinates, the center row coordinates, the length of the major axis, the length of the minor axis, and the direction angle of the major axis.
9. The system according to claim 6, characterized in that, The fixed marker points in the first module are designated fixed physical feature points on the rotating disk. The number of angle positions is at least five to ensure the reliability of ellipse fitting. The angle distribution is non-uniform and arbitrary. The image acquisition operation is controlled and executed by the robot end-side processor.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 9.