Monocular vision pose measurement method and system based on circular characteristic target spot

Abstract

The invention discloses a monocular vision pose measurement method and system based on a circular feature target spot, and the method comprises the following steps: camera calibration, image collection, image preprocessing, and feature point detection and matching: employing Hough circle detection to extract the center two-dimensional pixel coordinate and radius of the circular feature target spot, obtaining a uniquely determined relationship between the feature target spots according to the circle center coordinates and the radius information; performing pose measurement, and obtaining a rotation matrix R and a translation vector T which are converted from a target coordinate system in each collected pose image to a camera coordinate system according to a PnP algorithm; and solving the relative pose. According to the method, a specially designed cooperative target containing five circular characteristic target spots is used as a target object, and a rapid detection algorithm realized according to target characteristics is combined, so that circle center matching sorting is realized, and the precision is improved. The method has the advantages of being simple in structure, simple in camera calibration, large in measurement view field, high in real-time performance and the like, and meanwhile the defects that the view field is small and stereo matching is difficult in stereo vision are overcome.

Description

Technical Field

[0001] The present invention belongs to the technical field of visual detection, and in particular relates to a monocular visual posture measurement method and system based on a circular feature target. Background Art

[0002] An aviation turboshaft engine is an aircraft engine that outputs shaft power outward through a power output shaft. A coupling is a mechanical component that firmly connects the driving shaft and the driven shaft in different mechanisms so that they rotate together and transmit motion and torque. It is an important power transmission component of small and medium-sized aviation turboshaft engines and directly affects the power performance of the engine. At present, there are many types of couplings, which can be classified differently according to different principles. Generally speaking, couplings are divided into rigid couplings and flexible couplings. Unlike rigid couplings, flexible couplings have greater elasticity. In addition to being able to transmit torque to the driving shaft and the driven shaft, they are still tolerant of inconsistent motion in other degrees of freedom of the driving shaft and the driven shaft except for the axial rotational motion. Excessive stress will not accumulate to cause the coupling to break or be damaged. It is an important power transmission component of the engine.

[0003] Based on the principles of elastic mechanics, there is a one-to-one correspondence between the relative motion displacement and rotation at both ends of the flexible coupling and the load. Therefore, in addition to the axis rotation motion, the motion data of the other five degrees of freedom can characterize the load performance that the flexible coupling can accommodate during the torque transmission process of the active shaft and the driven shaft. It is an important performance indicator of the flexible coupling and requires corresponding measurement equipment for data collection and analysis. At the same time, although there are many types of existing posture measurement methods, most of the application objects are oriented to medium and large targets, such as the laser measurement equipment used by traditional methods in the real-time measurement of the terminal posture of industrial equipment such as machine tools and robots. However, the internal space of aircraft engines is small, and there is currently a lack of unified measurement methods for small objects in compact environments. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a monocular vision posture measurement method and system based on a circular feature target.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] The present invention provides a monocular vision posture measurement method based on a circular feature target, comprising the following steps:

[0007] Step S1: fix the target on the surface of the target to be measured, fix the monocular camera on a stationary platform corresponding to the target, obtain and record the parameters of the monocular camera, the target includes five highly reflective circular target points, one of which is located at the exact center of the target, and the other four highly reflective circular target points are located at the upper left, lower left, upper right, and lower right positions of the central target, respectively. The five highly reflective circular target points constitute a completely centrally symmetrical pattern, the four highly reflective circular target points around have the same diameter, the central highly reflective circular target point is used as circle No. 0, the upper left corner is circle No. 1, the upper right corner is circle No. 2, the lower right corner is circle No. 3, and the lower left corner is circle No. 4;

[0008] Step S2: controlling the lighting unit to illuminate the target, and using the camera to acquire an image sequence of the target;

[0009] Step S3: preprocessing the acquired image sequence;

[0010] Step S4: detecting, matching and sorting the characteristic target points of the preprocessed image sequence based on the target information to obtain the two-dimensional pixel coordinate information of all characteristic target points, specifically including: obtaining the two-dimensional pixel coordinates and radius information of the center of all circles in the image through the Hough circle detection algorithm, comparing the radii of all circles, taking the circle with the largest radius as the central circular target point, that is, circle No. 0, calculating the distance between the center of all circles and the origin of the pixel coordinate system, the circle with the smallest distance is taken as circle No. 1, and the circle with the largest distance is taken as circle No. 3, calculating the slope of the line connecting the center of all circles and the origin of the pixel coordinate system, the circle with the largest slope is taken as circle No. 4, and the circle with the smallest slope is taken as circle No. 2, and obtaining the two-dimensional pixel coordinate information and corresponding numbers of all characteristic target circles;

[0011] Step S5: obtaining the three-dimensional coordinate information of the characteristic target in the target coordinate system according to the target design parameters, and calculating the pose relationship of each image through the PnP algorithm according to the two-dimensional pixel coordinate information, the three-dimensional coordinate information, and the monocular camera parameters;

[0012] Step S6: According to the posture relationship of the images at adjacent moments, the relative posture results of the images at adjacent moments are calculated through matrix multiplication and matrix inversion operations.

[0013] Furthermore, the diameter of the central target is n times the diameter of the four surrounding highly reflective circular target points, where n is greater than 1.

[0014] Furthermore, in step S1, the monocular camera is fixed on a stationary platform corresponding to the target, wherein the target is perpendicular to the optical axis of the monocular camera.

[0015] Furthermore, in step S1, the monocular camera parameters are obtained and recorded, including the camera intrinsic parameter matrix K and the camera distortion coefficient D, wherein the camera intrinsic parameter matrix K is:

[0016]

[0017] Where f is the focal length of the camera, dx and dy are the intrinsic parameters of the camera sensor, representing the actual size of each pixel, u0 and v0 are the horizontal and vertical coordinates of the principal point, representing the offset of the origin of the image coordinate system relative to the origin of the pixel coordinate system;

[0018] The camera distortion coefficient D is:

[0019]

[0020] Among them, k1, k2 and k3 are the first-order, second-order and third-order coefficients of radial distortion, respectively, and p1 and p2 are the first-order and second-order coefficients of tangential distortion, respectively.

[0021] Furthermore, the lighting unit in step S2 is two LED lighting lamps, and the LED lighting lamps are installed on both sides of the monocular camera.

[0022] Furthermore, the preprocessing in step S3 includes: using camera intrinsic parameters and distortion coefficients to correct the image distortion, graying the corrected image, and filtering the grayscale image using Gaussian filtering.

[0023] Furthermore, the origin of the pixel coordinate system is the first pixel in the upper left corner of the image.

[0024] Further, the step S5 specifically includes: obtaining the three-dimensional coordinates of all circle centers in the target coordinate system according to the target design parameters, the origin of the target coordinate system is the center of the circle located at the central circular target point, the X axis of the coordinate system passes through the midpoint of the line connecting the centers of circles No. 2 and No. 3, the XOY plane is located on the upper surface of the target, and the Z axis is perpendicular to the upper surface of the target. The target design parameters are the center positions and radii of all circles, and the matrix P is formed according to the three-dimensional coordinates of all circle centers. o , the two-dimensional pixel coordinates of all circles obtained in step S4 form a matrix P u , the five characteristic target points correspond to the coordinate matrix P in the target coordinate system o , the coordinate matrix P of the 5 feature target points corresponding to the pixel coordinate system u , camera intrinsic parameter matrix K and camera distortion coefficient D are substituted into the formula:

[0025] [R,T]=PnP(P o ,P u ,K,D)

[0026] Obtain the pose relationship of the image: rotation matrix R, translation vector T.

[0027] Furthermore, the calculation process of step S6 includes: at time 1, the rotation matrix R1 and translation vector T1 between the camera coordinate system and the target coordinate system can be obtained by the PnP algorithm; at time 2, the rotation matrix R2 and translation vector T2 between the camera coordinate system and the target coordinate system can be obtained by the same method; the transformation matrix from the target coordinate system at time 1 to the camera coordinate system And the transformation matrix from the target coordinate system to the camera coordinate system at time 2

[0028]

[0029] There is a relationship between transformation matrices:

[0030]

[0031] According to the formula:

[0032]

[0033] The relative pose result can be obtained through matrix multiplication and matrix inversion operations, including the rotation matrix R and the translation vector T:

[0034] R=R1 · R2

[0035] T=R1 · (T2-T1).

[0036] The present invention also provides a monocular vision posture measurement system based on a circular feature target, comprising a monocular camera, two LED lighting units, a target and a processor, wherein the processor is connected to the monocular camera and is used to execute the above method.

[0037] Compared with the prior art, the present invention has the following advantages:

[0038] (1) The present invention uses a specially designed target as the target object, combined with a fast detection algorithm based on the target characteristics, to achieve center matching sorting and improve accuracy.

[0039] (2) The present invention has the advantages of simple structure, simple camera calibration, large measurement field of view, and strong real-time performance.

[0040] (3) The present invention avoids the shortcomings of small field of view and difficult stereo matching in stereo vision.

[0041] (4) The present invention has high measurement accuracy. The position and posture of the target object in the camera coordinate system can be obtained by using a specially designed target, which has high accuracy in small-scale measurement, and the measurement accuracy can reach 0.2mm. The specially designed target can be fixed on the measurement target and can be widely used in the measurement of modern industrial equipment (machine tools, industrial robot arms), meeting the needs of modern industrial equipment for small-scale high-precision posture measurement. Compared with traditional measurement equipment, this method requires low equipment cost and high posture measurement accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of the method of the present invention;

[0043] Figure 2 It is a system schematic diagram of the present invention;

[0044] Figure 3 A schematic diagram of a specially designed target of the present invention;

[0045] Figure 4 It is a flow chart of the detection matching and sorting algorithm of the present invention;

[0046] Figure 5 It is a coordinate system posture relationship diagram of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0048] Embodiment 1:

[0049] The present invention provides a monocular vision posture measurement method based on a circular feature target, such as Figure 1 As shown, including:

[0050] Step S1: Camera calibration before pose measurement.

[0051] Step S2: Acquire an image sequence of the cooperative target.

[0052] Step S3: pre-processing the collected image.

[0053] Step S4: Match and sort the five circular feature target points based on the circle center coordinates and radius information.

[0054] Step S5: The PnP algorithm is used to calculate the pose relationship between the target coordinate system and the camera coordinate system.

[0055] Step S6: Obtain the relative position and posture of the target coordinate systems in the image sequence from the Euclidean transformation relationship.

[0056] Furthermore, the monocular vision pose measurement method based on the circular feature target includes: S1: obtaining and recording the camera parameters. Before the visual measurement, the camera parameters are calibrated using a high-precision dot calibration plate and the camera intrinsic parameter matrix K is recorded:

[0057]

[0058] Where: f is the focal length of the camera, dx and dy are the intrinsic parameters of the camera sensor, representing the actual size of each pixel, u0 and v0 are the horizontal and vertical coordinates of the principal point, representing the offset of the origin of the image coordinate system relative to the origin of the pixel coordinate system; record the camera distortion coefficient D:

[0059]

[0060] Among them: k1, k2 and k3 are the first-order, second-order and third-order coefficients of radial distortion respectively, and p1 and p2 are the first-order and second-order coefficients of tangential distortion respectively.

[0061] S2: Use the camera to obtain the image of the cooperative target in real time. The cooperative target is specially designed and contains 5 circular characteristic targets. A highly reflective circular target is installed at the center, and 4 highly reflective circular targets are symmetrically and evenly distributed around it. The diameter of the central circular target is twice the diameter of the surrounding circular targets; the origin O of the target coordinate system is located at the center of the central circular target, the XOY plane is located on the upper surface of the target, and the Z axis is perpendicular to the upper surface of the target.

[0062] S3: Preprocess the collected image. Use the camera calibration parameters to correct the image distortion; then grayscale the image; finally use the filter to filter the image.

[0063] S4: Match and sort the five circular feature targets based on the coordinates of the circle center and the radius information. Detect all circles in the cooperative target image, obtain the two-dimensional pixel coordinates and radius of all circles, identify the central circular target with the largest radius as circle 0; identify the surrounding circular target with the smallest distance between the circle center and the origin of the pixel coordinate system as circle 1; identify the surrounding circular target with the largest distance between the circle center and the origin of the pixel coordinate system as circle 3; identify the surrounding circular target with the smallest slope of the line connecting the circle center and the origin of the pixel coordinate system as circle 2; identify the surrounding circular target with the largest slope of the line connecting the circle center and the origin of the pixel coordinate system as circle 4; match and sort the five circles in the order of 0 to 4.

[0064] S5: The PnP algorithm is used to calculate the pose relationship between the target coordinate system and the camera coordinate system. Using the PnP algorithm in visual measurement, the five characteristic target points on the cooperative target correspond to the coordinate matrix P in the target coordinate system. o , the coordinate matrix P of the 5 feature target points corresponding to the pixel coordinate system u , the camera intrinsic parameter matrix K and the camera distortion coefficient D are used as the PnP algorithm parameter formula (1) to obtain the position and posture relationship between the target coordinate system and the camera coordinate system in a single image: rotation matrix (R) and translation vector (T).

[0065] [R,T]=PnP(P o ,P u ,K,D) (1).

[0066] S6: Obtain the relative pose of the target coordinate system in the image sequence by using the Euclidean transformation relationship. According to the pose relationship of the target coordinate system in a single image relative to the camera coordinate system, the relative pose relationship between the target coordinate systems in the two images is obtained by using the Euclidean transformation principle.

[0067] Embodiment 2:

[0068] The parts not mentioned in this embodiment are the same as those in Embodiment 1.

[0069] The system diagram of this embodiment is as follows Figure 2 As shown in the figure, the target is fixed on the surface of the target to be measured, and the monocular camera is fixed on a static platform. The LED lighting unit installed around the camera is controlled to illuminate the target, and the five highly reflective marking points installed on the target surface reflect the incident light back and then enter the camera field of view for imaging. When the target to be measured moves, it drives the target to move. The camera takes real-time photos of the target to obtain the target image at the same time interval. After feature processing of the image, the effective feature point information is extracted, and then the pose solution algorithm is used to obtain the pose information of the target.

[0070] The monocular camera is the MERCURY2 digital camera of Daheng Imaging Co., Ltd., model MERCURY2-502-79U3M, with a resolution of 2448×2048, a USB3.0 data interface, and can be powered by USB; the lens is the HN-P-20M series fixed-focus lens of Daheng Imaging Co., Ltd., model HN-P-2524-20M-C1.1 / 1, with a focal length of 25mm and a field of view of 37.31°×30.47°×22.53°. The lens and camera are connected through a C-type interface.

[0071] The proposed method is implemented in the programming language C++ on the Windows operating system, in which some data structures and common image processing algorithms of OpenCV are also used. The software environment is Visual Studio 2017 and OpenCVv4.5.1.

[0072] The hardware environment is Intel i7-7700HQ processor and 16GB running memory.

[0073] This example uses a specially designed target, the target schematic diagram is as follows Figure 3 The diameter of the central circular target is 16 mm, the diameter of the surrounding four circular targets is 8 mm, and the distance between adjacent ones is 34 mm. The five circular targets form a completely centrally symmetrical pattern. Figure 3 The XYZ coordinate system in is the target coordinate system. The origin O of the coordinate system is located at the center of the central circular target. The X axis of the coordinate system passes through the midpoint of the line connecting the centers of circles 2 and 3. The XOY plane is located on the upper surface of the target, and the Z axis is perpendicular to the upper surface of the target.

[0074] S1: Before performing posture measurement, the camera is calibrated using the high-precision dot calibration plate shown in Table 1 according to the Zhang Zhengyou camera calibration method.

[0075] Table 1 Dot calibration plate related parameters

[0076]

[0077] The camera internal parameter matrix is ​​obtained as:

[0078]

[0079] The camera distortion coefficient is:

[0080]

[0081] S2: Use the camera to obtain an image containing the target. The target should be as perpendicular to the camera optical axis as possible and located where the camera can image clearly. The image should occupy more than 1 / 5 of the entire frame.

[0082] S3: First, use the camera intrinsic parameters and distortion coefficients to dedistort the image; then grayscale the corrected image; finally, use Gaussian filtering to filter the grayscale image. According to the test, the Gaussian filter size is selected as (9, 9), the standard deviation of the Gaussian filter in the X direction is selected as 2, and the standard deviation of the Gaussian filter in the Y direction is selected as 2. At this time, the noise removal effect and the clarity of the filtered image are more balanced, and the filtering effect is better.

[0083] S4: Detect, match and sort the five circular feature targets in the preprocessed image. See the process for details. Figure 4. Detect and extract the center coordinates (u, v) and radius information of all circles in the image. According to the test, the inverse ratio of the resolution to the image resolution during discretization is selected as 1, the minimum distance between the two center points in the detection result is selected as 300, the larger value of the two thresholds passed to the Canny edge detector is selected as 100, the accumulator threshold for detecting circles is selected as 20, the minimum radius of the detected circle is selected as 50, and the maximum radius of the detected circle is selected as 160. At this time, 5 circular feature targets can be accurately detected.

[0084] By comparing the radii of all circles, the circle with the largest radius is selected, and this circle is number 0.

[0085] In the acquired image, the origin of the pixel coordinate system is the first pixel in the upper left corner of the image. The distance between the extracted centers and the origin of the pixel coordinate system is calculated. The distance calculation formula is as follows:

[0086]

[0087] The one with the largest distance is No. 3, and the one with the smallest distance is No. 1.

[0088] Calculate the slope of the line connecting all the circle centers and the origin of the pixel coordinate system. The slope calculation formula is as follows:

[0089] k=u / v (2)

[0090] Among them, No. 4 has the largest slope, and No. 2 has the smallest slope.

[0091] Since the target rotates around the Z axis at a very small angle in actual measurement, the corresponding relationship between the two-dimensional center of the circle after detection and sorting in the image and the three-dimensional center of the circle in the target coordinate system is uniquely determined.

[0092] S5: The XOY plane of the target coordinate system is located on the upper surface of the target. For the three-dimensional coordinates (X, Y, Z) of the centers of the five circular feature target points, Z = 0. The coordinate arrays of the target coordinate system of the five centers of the target from No. 0 to No. 4 can be obtained from the target design parameters, which are: (0, 0, 0), (-17, 17, 0), (17, 17, 0), (17, -17, 0), (-17, -17, 0), arranged in order into a matrix P with 5 rows and 3 columns o .

[0093] According to the detection matching sorting algorithm, the coordinate array of the pixel coordinate system of the five circle centers from 0 to 4 can be obtained, which is arranged in sequence into a matrix P with 5 rows and 2 columns. u .

[0094] Using the PnP (Perspective-n-Point) algorithm in visual measurement, the five characteristic target points correspond to the coordinate matrix P in the target coordinate system. o, the coordinate matrix P of the 5 feature target points corresponding to the pixel coordinate system u , the camera intrinsic parameter matrix K and the camera distortion coefficient D are substituted into formula (3) to obtain the pose relationship of the target coordinate system in a single image transformed to the camera coordinate system: rotation matrix (R) and translation vector (T).

[0095] [R,T]=PnP(P o ,P u ,K,D) (3).

[0096] S6: Figure 5 As shown, at time 1, the target coordinate system is O1-X1Y1Z1, and at time 2 after the target moves, its coordinate system becomes O2-X2Y2Z2, and the camera coordinate system is O c -X c Y c Z c .

[0097] At time 1, the PnP algorithm can be used to calculate the rotation matrix R1 and translation vector T1 between the camera coordinate system and the target coordinate system.

[0098] At time 2, the rotation matrix R2 and translation vector T2 between the camera coordinate system and the target coordinate system can be calculated by the same method.

[0099] Then the transformation matrix of the target coordinate system from time 2 to time 1 is It can be expressed as:

[0100]

[0101] Similarly, the transformation matrix from the target coordinate system at time 1 to the camera coordinate system can be obtained: And the transformation matrix from the target coordinate system to the camera coordinate system at time 2

[0102]

[0103] There is a relationship between transformation matrices:

[0104]

[0105] According to formula (7), we have:

[0106]

[0107] The relative pose result can be obtained through matrix multiplication and matrix inversion operations, including the rotation matrix R and the translation vector T:

[0108] R=R1 · R2 (9)

[0109] T=R1· (T2-T1) (10)

[0110] An experimental platform was built to verify and test the monocular vision pose measurement system and method based on circular feature targets.

[0111] The translation range is set to ±10mm, and the machine moves along the X, Y, and Z axes in steps of 1mm. It pauses for a few seconds at each position, and the image in the current position is collected by the monocular camera. The rotation range is set to ±5°, and the machine rotates along the X, Y, and Z axes in steps of 1°. It pauses for a few seconds at each position, and the image in the current position is collected by the monocular camera.

[0112] The measured value is obtained by image solution, and the measurement error can be obtained by comparing the theoretical value with the measured value.

[0113] The error indicators in each direction are shown in Table 2 and Table 3.

[0114] Table 2 Translation error index

[0115]

[0116] Table 3 Rotation error index

[0117]

[0118] The experimental results show that the translational measurement accuracy of this system example is better than 0.2mm, and the rotational measurement accuracy is better than 0.2°. The measurement accuracy meets actual needs.

[0119] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A monocular vision pose measurement method based on a circular feature target, characterized in that: The following steps are involved: Step S1: fix the target on the surface of the target to be measured, fix the monocular camera on a stationary platform corresponding to the target, obtain and record the parameters of the monocular camera, the target includes five highly reflective circular target points, one of which is located at the exact center of the target, and the other four highly reflective circular target points are located at the upper left, lower left, upper right, and lower right positions of the central target, respectively. The five highly reflective circular target points constitute a completely centrally symmetrical pattern, the four highly reflective circular target points around have the same diameter, the central highly reflective circular target point is used as circle No. 0, the upper left corner is circle No. 1, the upper right corner is circle No. 2, the lower right corner is circle No. 3, and the lower left corner is circle No. 4; Step S2: controlling the lighting unit to illuminate the target, and using the camera to acquire an image sequence of the target; Step S3: preprocessing the acquired image sequence; Step S4: detecting, matching and sorting the characteristic target points of the preprocessed image sequence based on the target information to obtain the two-dimensional pixel coordinate information of all characteristic target points, specifically including: obtaining the two-dimensional pixel coordinates and radius information of the center of all circles in the image through the Hough circle detection algorithm, comparing the radii of all circles, taking the circle with the largest radius as the central circular target point, that is, circle No. 0, calculating the distance between the center of all circles and the origin of the pixel coordinate system, the circle with the smallest distance is taken as circle No. 1, and the circle with the largest distance is taken as circle No. 3, calculating the slope of the line connecting the center of all circles and the origin of the pixel coordinate system, the circle with the largest slope is taken as circle No. 4, and the circle with the smallest slope is taken as circle No. 2, and obtaining the two-dimensional pixel coordinate information and corresponding numbers of all characteristic target circles; Step S5: obtaining the three-dimensional coordinate information of the characteristic target in the target coordinate system according to the target design parameters, and calculating the pose relationship of each image through the PnP algorithm according to the two-dimensional pixel coordinate information, the three-dimensional coordinate information, and the monocular camera parameters; Step S6: According to the posture relationship of the images at adjacent moments, the relative posture results of the images at adjacent moments are calculated through matrix multiplication and matrix inversion operations.

2. The monocular vision posture measurement method based on a circular feature target according to claim 1 is characterized in that: The diameter of the central target is n times the diameter of the four surrounding highly reflective circular target points, where n is greater than 1.

3. The monocular vision posture measurement method based on a circular feature target according to claim 1 is characterized in that: In the step S1, the monocular camera is fixed on a stationary platform corresponding to the target, wherein the target is perpendicular to the optical axis of the monocular camera.

4. The monocular vision posture measurement method based on a circular feature target according to claim 1 is characterized in that: In step S1, the monocular camera parameters are obtained and recorded, including the camera intrinsic parameter matrix K and the camera distortion coefficient D, where the camera intrinsic parameter matrix K is: Where f is the focal length of the camera, dx and dy are the intrinsic parameters of the camera sensor, representing the actual size of each pixel, u0 and v0 are the horizontal and vertical coordinates of the principal point, representing the offset of the origin of the image coordinate system relative to the origin of the pixel coordinate system; The camera distortion coefficient D is: Among them, k1, k2 and k3 are the first-order, second-order and third-order coefficients of radial distortion, respectively, and p1 and p2 are the first-order and second-order coefficients of tangential distortion, respectively.

5. The monocular vision posture measurement method based on a circular feature target according to claim 1, characterized in that: The lighting unit in step S2 is two LED lighting lamps, and the LED lighting lamps are installed on both sides of the monocular camera.

6. The monocular vision posture measurement method based on a circular feature target according to claim 1 is characterized in that: The preprocessing in step S3 includes: using camera internal parameters and distortion coefficients to correct the image distortion, graying the corrected image, and filtering the grayscale image using Gaussian filtering.

7. The monocular vision posture measurement method based on a circular feature target according to claim 1 is characterized in that: The origin of the pixel coordinate system is the first pixel in the upper left corner of the image.

8. The monocular vision posture measurement method based on a circular feature target according to claim 1, characterized in that: The step S5 specifically includes: obtaining the three-dimensional coordinates of all circle centers in the target coordinate system according to the target design parameters, wherein the origin of the target coordinate system is the center of the circle located at the central circular target point, the X axis of the coordinate system passes through the midpoint of the line connecting the centers of circles 2 and 3, the XOY plane is located on the upper surface of the target, and the Z axis is perpendicular to the upper surface of the target. The target design parameters are the center positions and radii of all circles, and the matrix P is formed according to the three-dimensional coordinates of all circle centers. o , the two-dimensional pixel coordinates of all circles obtained in step S4 form a matrix P u , the five characteristic target points correspond to the coordinate matrix P in the target coordinate system o , the coordinate matrix P of the 5 feature targets corresponding to the pixel coordinate system u , camera intrinsic parameter matrix K and camera distortion coefficient D are substituted into the formula: [R,T]=PnP(P o ,P u ,K,D) Obtain the pose relationship of the image: rotation matrix R, translation vector T.

9. The monocular vision posture measurement method based on a circular feature target according to claim 1, characterized in that: The calculation process of step S6 includes: at time 1, the rotation matrix R1 and translation vector T1 between the camera coordinate system and the target coordinate system can be obtained by the PnP algorithm; at time 2, the rotation matrix R2 and translation vector T2 between the camera coordinate system and the target coordinate system can be obtained by the same method; the transformation matrix from the target coordinate system at time 1 to the camera coordinate system And the transformation matrix from the target coordinate system to the camera coordinate system at time 2 There is a relationship between transformation matrices: According to the formula: The relative pose result can be obtained through matrix multiplication and matrix inversion operations, including the rotation matrix R and the translation vector T: R=R1 · R2 T=R1 · (T2-T1)。 10. A monocular vision posture measurement system based on a circular feature target, characterized in that: The invention comprises a monocular camera, two LED lighting units, a target and a processor, wherein the processor is connected to the monocular camera and is used to execute the methods of claims 1 to 9.

Images